CN104572952B - The recognition methods of live multimedia file and device - Google Patents

Abstract

The invention discloses a method and a device for identifying live multimedia files. The method comprises: obtaining the feature information of the current period of the live multimedia file according to the real-time data stream of the input live multimedia file; locating the multimedia record to be updated in the feature database according to the identification information of the live multimedia file; Information update the feature sample in the multimedia record to be updated; receive the identification request for identifying the target multimedia file, match the feature information of the target multimedia file included in the identification request with the feature sample in the feature database, to locate the multimedia record corresponding to the target multimedia file ; Obtain the identification information of the multimedia file corresponding to the target multimedia file. Through the present invention, live video can be identified.

Description

Method and device for identifying live multimedia files

technical field

The present invention relates to the technical field of video identification, in particular to a method and device for identifying live multimedia files.

Background technique

The current video search method usually uses the "keyword" of the video to search. This not only requires the user to know the relevant information of the video, but also requires the search service provider to maintain a "keyword" database corresponding to the video in a timely manner. In fact, we often encounter such embarrassment: we come across an interesting video in the streets or in front of the TV, but we are not familiar with or even know the information of this video, let alone search through "keywords" It's time for this video.

Therefore, video recognition based on sound came into being under the impetus of this practical demand. In the technology based on sound recognition video, when the user needs to identify a certain video, first obtain the sound information in the video through the recording device of the mobile terminal (such as a smart phone), and compare the feature data reflecting the sound information with the cloud server. The feature database performs matching, and returns the matching result (video stream or video-related information) to the mobile terminal.

However, video files have the characteristics of fast updating and fast online, and even many video files are in the form of webcasting, so the video that users need to identify is often the video that is being broadcast. However, in the above method of the prior art, the cloud server will construct a feature database according to the audio information corresponding to the video after the cloud server obtains the complete video generated by the video source. Therefore, the method of the prior art cannot identify the live video.

Aiming at the problem that the existing technology cannot recognize the live video, no effective solution has been proposed yet.

Contents of the invention

The main purpose of the present invention is to provide a method and device for identifying live multimedia files, so as to solve the problem that the prior art cannot identify live video.

According to one aspect of the present invention, a method for identifying live multimedia files is provided.

The identification method of the live multimedia file according to the present invention comprises: according to the real-time data stream of the input live multimedia file, obtains the feature information of the current period of the live multimedia file; according to the identification information of the live multimedia file, locate the multimedia record to be updated in the feature database, Wherein, the feature database is used to store at least one multimedia record, the multimedia record includes feature samples of multimedia files, identification information corresponding to feature samples, and the time length of feature samples is the first predetermined time; update according to feature information of the current period of live multimedia files The feature sample in the multimedia record to be updated; receive the recognition request for identifying the target multimedia file, match the feature information of the target multimedia file included in the recognition request and the feature sample in the feature database, to locate the corresponding multimedia record of the target multimedia file; obtain Identification information of the multimedia file corresponding to the target multimedia file.

Further, the characteristic information is the fingerprint information of the audio data of the multimedia file, and obtaining the characteristic information of the current period of the live multimedia file according to the input real-time data stream of the live multimedia file includes: obtaining the audio frequency of the current period of the live multimedia file according to the real-time data stream Data; the audio data of the current period is divided into a plurality of audio segments of the second predetermined time in chronological order, wherein the second predetermined time is less than the first predetermined time; and the fingerprint information of each audio segment is extracted to obtain the live multimedia Feature information for the current period.

Further, the feature sample is the fingerprint information of n audio clips, the feature information of the current period of the live multimedia file is the fingerprint information of m audio clips, m<n, the time length of n audio clips is the first predetermined time, according to The feature information of the live multimedia file updates the feature samples in the multimedia record to be updated including: deleting the earliest m fingerprint information of the feature sample in the multimedia record to be updated; the m fingerprint information of the current period of the live multimedia file in time order Placed in the feature sample of the multimedia record to be updated.

Further, update the feature sample in the multimedia record to be updated according to the feature information of the current period of the live multimedia file, specifically including: Step S1: the feature pointer points to the first fingerprint information in the feature information of the current period of the live multimedia file, and The timer is cleared to start feature extraction timing; Step S2: Obtain the fingerprint information pointed to by the feature pointer; Step S3: Extract the feature sample of the multimedia record corresponding to the identification information of the live multimedia to obtain the first feature sample; Step S4: The fingerprint information pointed to by the feature pointer is spliced to the end of the first feature sample to obtain the second feature sample; step S5: delete a fingerprint information from the beginning of the second feature sample; step S6: judge whether the time in the timer reaches the first 3. Scheduled time, if the third scheduled time is not reached, the feature pointer points to the next fingerprint information, and steps S2 to S6 are repeated; if the third scheduled time is reached, the obtained second feature sample is used to replace the multimedia identification corresponding to the multimedia record The feature sample, wherein the third predetermined time is the playing time of the multimedia file corresponding to the m fingerprint information.

Further, extracting the fingerprint information of the audio clip includes: merging the left channel data and the right channel data of the audio clip to obtain the stereo data of the audio clip; and extracting the time-frequency characteristic data of the stereo data of the audio clip as the fingerprint of the audio clip information.

Further, the feature information of the target multimedia file included in the identification request is N fingerprint information of the current period of the live multimedia file, and one fingerprint information in the N fingerprint information is one of the N stereo data of the target multimedia. Time-frequency feature data, wherein, the i-th stereo data in the N stereo data is si'=ai'*l'+bi'*r', ai'+bi'=1, and l' is the current of the live multimedia file The left channel data of the period, r' is the right channel data of the current period of the live multimedia file, ai' and bi' are preset parameters, i=1, 2, 3...N. In the method, matching the feature information of the target multimedia file included in the identification request with the feature samples in the feature database to locate the multimedia record corresponding to the target multimedia file includes: comparing each fingerprint information of the target multimedia file with the feature sample in the feature database match the feature samples to obtain the matching rate of each fingerprint information; the multimedia record where the feature sample corresponding to the maximum matching rate is located is used as the multimedia record corresponding to the target multimedia file.

According to one aspect of the present invention, a device for identifying live multimedia files is provided.

The identification device of live multimedia file according to the present invention comprises: acquisition module, is used for obtaining the characteristic information of live multimedia file current period according to the real-time data stream of the live multimedia file of input; Locate the multimedia record to be updated in the feature database, wherein the feature database is used to store at least one multimedia record, the multimedia record includes a feature sample of the multimedia file, identification information corresponding to the feature sample, and the time length of the feature sample is the first predetermined time; The update module is used to update the feature sample in the multimedia record to be updated according to the feature information of the current period of the live multimedia file; the matching module is used to receive the identification request for identifying the target multimedia file, and match the feature of the target multimedia file included in the identification request The feature sample in the information and feature database is used to locate the multimedia record corresponding to the target multimedia file; the identification module is used to obtain the identification information of the multimedia file corresponding to the target multimedia file.

Further, the feature information is the fingerprint information of the audio data of the multimedia file, and the acquisition module includes: an audio data acquisition module, which is used to obtain the audio data of the current period of the live multimedia file according to the real-time data stream; The audio data of the time period is divided into a plurality of audio segments of the second predetermined time in chronological order, wherein the second predetermined time is less than the first predetermined time; and the fingerprint information extraction module is used to extract the fingerprint information of each audio segment to obtain Feature information of the current period of live multimedia.

Further, the feature sample is the fingerprint information of n audio clips, the feature information of the current period of the live multimedia file is the fingerprint information of m audio clips, m<n, the time length of n audio clips is the first predetermined time, update The module includes: a deletion module, which is used to delete the earliest m fingerprint information of the feature sample in the multimedia record to be updated; an addition module, which is used to place the m fingerprint information of the current period of the live multimedia file in chronological order in the to-be-updated Among the feature samples of multimedia records.

Further, the update module specifically performs the following steps:

Step S1: The feature pointer points to the first fingerprint information in the feature information of the current period of the live multimedia file, and the timer is reset to start feature extraction timing; Step S2: Obtain the fingerprint information pointed to by the feature pointer; Step S3: Extract and live broadcast The feature sample of the multimedia record corresponding to the multimedia identification information to obtain the first feature sample; Step S4: splicing the fingerprint information pointed to by the feature pointer to the end of the first feature sample to obtain the second feature sample; Step S5: From The beginning of the second feature sample deletes a fingerprint information; step S6: judge whether the time in the timer reaches the third predetermined time, if it does not reach the third predetermined time, the feature pointer points to the next fingerprint information, and repeat steps S2 to S6: If the third predetermined time is reached, replace the characteristic sample corresponding to the multimedia identifier in the multimedia record with the obtained second characteristic sample, wherein the third predetermined time is the playing time of the multimedia file corresponding to the m fingerprint information.

Further, the fingerprint information extraction module includes: a stereo data synthesis module, which is used to merge the left channel data and the right channel data of the audio clip to obtain the stereo data of the audio clip; and a time-frequency feature extraction module, which is used to extract the audio clip The time-frequency feature data of the stereo data is used as the fingerprint information of the audio segment.

Further, the feature information of the target multimedia file included in the identification request is N fingerprint information of the current period of the live multimedia file, and one fingerprint information in the N fingerprint information is one of the N stereo data of the target multimedia. Time-frequency feature data, wherein, the i-th stereo data in the N stereo data is si'=ai'*l'+bi'*r', ai'+bi'=1, and l' is the current of the live multimedia file The left channel data of the period, r' is the right channel data of the current period of the live multimedia file, ai' and bi' are preset parameters, i=1,2,3...N, in this device, the matching module Including: a matching rate determination module, which is used to match each fingerprint information of the target multimedia file with the feature samples in the feature database to obtain the matching rate of each fingerprint information; a multimedia record determination module, which is used to match the maximum matching rate corresponding to The multimedia record where the feature sample is located serves as the multimedia record corresponding to the target multimedia file.

Through the present invention, a feature database is preset to store feature information of live multimedia, specifically, at least one multimedia record is stored in the feature database, the multimedia record includes feature samples of multimedia files, identification information corresponding to feature samples, and feature samples The length of time is the first predetermined time. When there is a real-time data stream input of a live multimedia file, first obtain the feature information of the current period of the live multimedia file according to the input real-time data stream of the live multimedia file, and then according to the identification information of the live multimedia file Locate the multimedia record to be updated in the feature database, and update the feature samples in the multimedia record to be updated according to the feature information of the current period of the live multimedia file, thereby ensuring that the latest feature information of the live multimedia file is stored in the feature database. When receiving the identification request for identifying the target multimedia file, match the feature information of the target multimedia file included in the identification request with the feature samples in the feature database to locate the multimedia record corresponding to the target multimedia file, and then obtain the multimedia file corresponding to the target multimedia file identification information, so as to achieve the purpose of identifying the target multimedia file, and solve the problem that the live video cannot be identified in the prior art.

The above description is only an overview of the technical solution of the present invention. In order to better understand the technical means of the present invention, it can be implemented according to the contents of the description, and in order to make the above and other purposes, features and advantages of the present invention more obvious and understandable , the specific embodiments of the present invention are enumerated below.

Description of drawings

Various other advantages and benefits will become apparent to those of ordinary skill in the art upon reading the following detailed description of the preferred embodiment. The drawings are only for the purpose of illustrating a preferred embodiment and are not to be considered as limiting the invention. Also throughout the drawings, the same reference numerals are used to designate the same parts. In the attached picture:

Fig. 1 is a method flowchart according to Embodiment 1 of the present invention;

FIG. 2 is a flow chart of a method according to Embodiment 2 of the present invention;

FIG. 3 is a flow chart of a method according to Embodiment 3 of the present invention;

FIG. 4 is a schematic diagram of a system according to Embodiment 4 of the present invention;

FIG. 5 is a block diagram of a terminal according to Embodiment 4 of the present invention;

6 is a block diagram of a video retrieval server according to Embodiment 4 of the present invention;

Fig. 7 is a block diagram of a fingerprint management server according to Embodiment 4 of the present invention;

8 is a block diagram of a video management server according to Embodiment 4 of the present invention; and

Fig. 9 is a block diagram of a device according to Embodiment 5 of the present invention.

Detailed ways

The present invention will be further described below in conjunction with the accompanying drawings and specific embodiments. It should be pointed out that, in the case of no conflict, the embodiments in the present application and the features in the embodiments can be combined with each other.

The embodiment of the present invention provides a method for identifying a live multimedia file. In the method, a feature database is preset. During the live broadcast of the live multimedia file, the feature database is updated according to the real-time data stream of the live multimedia file to ensure that the feature database is stored in the feature database. There is the latest information on the current live multimedia files. When the user needs to identify the target multimedia file of the live broadcast, the feature information of the target multimedia file is matched with the feature samples in the preset feature database. If the feature information of the target multimedia file matches a certain feature sample successfully, the feature sample is passed. The corresponding identification information can achieve the purpose of identifying the target multimedia file.

It should be noted that there is usually a time difference between the live broadcast time of the live multimedia file and the generation time of the real-time data stream, and the live broadcast time of the live multimedia file (that is, the time when it is delivered to the user) is later than the generation time of the real-time data stream of the live multimedia file , the embodiment of the present invention is based on this time difference, so that the update time of the feature sample in the feature database is earlier than or synchronized with the live broadcast time of the live multimedia file, so that the feature database can keep storing the feature information of the current live multimedia file, and then can be in The live target multimedia file is matched to the feature database to achieve the purpose of identifying the target multimedia file.

Specifically, one or more multimedia records are stored in the feature database, and each multimedia record corresponds to a live multimedia file. For example, if the multimedia file is a video, each multimedia record corresponds to a live video; the multimedia file is audio, Then each multimedia record corresponds to a live audio. Each multimedia record includes a feature sample of a live multimedia file and identification information of the live multimedia file, wherein the feature sample is composed of feature information of a multimedia file with a preset fixed time length, for example, the feature sample is a feature of a video with a fixed time length Information composition; identification information is information that can identify different live multimedia.

When there is a real-time data stream input of a live multimedia file, the feature information of the current period of the live multimedia file is further obtained according to the real-time data stream, and the multimedia record corresponding to the live multimedia file is found in the feature database by the identification information of the live multimedia file, Then, the characteristic samples in the multimedia record are updated by the characteristic information, so as to ensure that the latest characteristic information of the live multimedia file is stored in the characteristic database.

When receiving the identification request for identifying the target multimedia file, further obtain the feature information of the target multimedia file according to the identification request, then match the feature information of the target multimedia file with the feature sample in the feature database, and locate the location by the matched feature sample A multimedia record, and then obtain the identification information in the multimedia record, and finally obtain the identification result of the target multimedia file from the identification information.

Any identification method in the embodiments of the present invention can be used in a search method for live multimedia files. In the search method for live multimedia files, identify the target multimedia file through the identification method of the live multimedia file in the embodiment of the present invention, that is, after obtaining the identification information of the target multimedia file, find the link of the target multimedia file through the identification information, and then search to the target multimedia file.

For example, a mobile phone user sees a live video on a street advertisement screen, and hopes to search for and play the video through the mobile phone. The sound data generates a target video recognition request and sends it to the cloud server. After the cloud server recognizes the target video using the recognition method of the embodiment of the present invention, in one case the cloud server can return the identification information of the target video to the mobile terminal, and the mobile terminal can then use the target video to identify the target video. The identification information of the video finds the link of the target video, and then plays the video through the found link; in another case, the cloud server can find the link of the target video through the identification information of the target video and return it to the mobile terminal, and then the mobile terminal passes the link. Link to play video.

Various embodiments provided by the present invention will be described in detail below.

Embodiment one

The first embodiment provides an embodiment of a method for identifying a live multimedia file. The execution body of the method provided in this embodiment is a cloud server, and the user terminal sends a target multimedia file identification request. Wherein, a feature database is preset in the cloud server, and the latest information of the current live multimedia file is stored in the feature database. When the user terminal needs to identify the target multimedia file of the live broadcast, the cloud server receives the identification request, matches the characteristic information of the target multimedia file with the characteristic samples in the preset characteristic database, and identifies the target through the identification information corresponding to the successfully matched characteristic samples multimedia files.

Fig. 1 is a flow chart of a method according to Embodiment 1 of the present invention. As shown in Fig. 1, the method specifically includes the following steps S102 to S116, wherein, steps S102 to S108 realize updating of the feature database, and steps S110 to S116 realize A target multimedia file is identified based on the feature database.

Step S102: Obtain the real-time data stream and identification information of the live multimedia file.

The cloud server communicates with the background database server of the live broadcast source to obtain the real-time data stream of the live multimedia file in real time. While obtaining the real-time data stream, the identification information of the live multimedia file can also be obtained. The multimedia file here can be video or audio, and the obtained real-time data stream is correspondingly video stream or audio stream. The identification information of the live multimedia file is information that can uniquely identify and determine the live multimedia file among multiple multimedia files (including live multimedia files and non-live multimedia files).

For live multimedia files, a live broadcast source can only live broadcast one multimedia file at a time, and the identity information of the live broadcast source has the characteristics of simplicity, uniqueness, and high recognition. Therefore, the identification information of the live multimedia file is preferably the live source. Identity information, for example, when the live multimedia file is a video, its identification information is the channel data of the video source that plays the live video, specifically, if the live multimedia file is a news broadcast, its identification information is "CCTV1"; another example, the live multimedia file is For audio, its identification information is the channel data of the audio source that plays the live audio. Specifically, for example, the live multimedia file is a live storytelling radio series, and its identification information is "Central People's Broadcasting Station".

Step S104: Obtain feature information of the current period of the live multimedia file according to the real-time data stream.

After each time the cloud server obtains the real-time data stream of the live multimedia file, it uses the preset feature extraction module to process the real-time data stream to extract the feature data of the real-time data stream and obtain the feature information of the current period of the live multimedia file.

For example, the video stream data of the live video is obtained, and the feature extraction module is used to process the video stream data to extract the corresponding audio fingerprint of the live video, and obtain the feature information of the current period of the live video.

Step S106: Locate the multimedia record to be updated in the feature database according to the identification information of the live multimedia file.

Wherein, the feature database stores a plurality of multimedia records, and each multimedia record includes two parts: a feature sample of the multimedia file and identification information of the multimedia file corresponding to the feature sample, wherein the time length of the feature sample is fixed and is updated each time Afterwards, the feature information in the feature sample is correspondingly changed to the updated feature information, but the time length of the feature sample does not change, so as to ensure that the feature sample stored in the feature database is always the feature information of the latest period of live multimedia files.

After obtaining the identification information of the live multimedia file in step S102, in this step S106, search the characteristic database, can locate the multimedia record corresponding to the identification information of the live multimedia file in the characteristic database, this multimedia record is to be updated multimedia recording.

Step S108: Update the feature samples in the multimedia record to be updated according to the feature information of the live multimedia file.

After the multimedia record to be updated is located, feature samples in the multimedia record are updated using feature information of the live multimedia file. When updating, when the time length of the feature information of the current period of the live multimedia file is greater than or equal to the time length of the feature sample, all the feature information in the feature sample can be updated; when the time length of the feature information of the current period of the live multimedia file is less than the feature sample Partial updates can be made when the time length of the sample is specified.

No matter what update method is adopted, in order to avoid data conflicts, the current feature database can be backed up, the backed up feature database can be updated, and then the original feature database can be overwritten with the updated backup feature database.

In the cloud server, the data streams of different live multimedia files are received in real time, and the feature database is updated in time through steps S102 to S108, so that for any live multimedia file, the feature information of the latest period of time is always stored in the feature database .

Step S110: Receive an identification request for identifying a target multimedia file.

The identification request can be sent by the user terminal. When the user needs to identify a multimedia file that is being broadcast live (this application defines the multimedia file that is to be identified as the target multimedia file), the data stream of the target multimedia file within a certain period of time is obtained. , when the amount of data stream data in this period of time is large, the feature information of the target multimedia file can be extracted at the user terminal, and the extracted feature information can be packaged as a recognition request and sent to the cloud server, and the cloud server receives the target multimedia file. Identification request of feature information to achieve the purpose of reducing the amount of data transmission; when the data volume of the data stream is small, the data stream can be directly encapsulated as an identification request and sent to the cloud server, and the cloud server receives the identification of the data stream containing the target multimedia file request to reduce the requirements on the data processing capability of the user terminal.

Step S112: Obtain feature information of the target multimedia file according to the identification request.

After the cloud server receives the identification request, it obtains the characteristic information according to the identification request. Specifically, when the identification request contains the characteristic information of the target multimedia file, the cloud server can obtain the characteristic information of the target multimedia file by parsing the identification request; when the identification request contains For the data stream of the target multimedia file, the cloud server invokes a preset feature extraction module to perform data processing on the data stream of the target multimedia file, so as to extract feature information of the target multimedia file.

Step S114: Match the feature information of the target multimedia file with feature samples in the feature database to locate the multimedia record corresponding to the target multimedia file.

After the cloud server obtains the feature information of the target multimedia file, it matches with each feature sample in the feature database. If the feature information of the target multimedia file matches a certain feature sample successfully, the matching ends, and the multimedia record where the feature sample is located is It is the multimedia record corresponding to the target multimedia file.

Step S116: Obtain the identification information in the multimedia record corresponding to the target multimedia file, so as to identify the target multimedia file.

After the multimedia record corresponding to the target multimedia file is located, the purpose of identifying the target multimedia file can be achieved through the identification information in the multimedia record. For example, if the identification information in the multimedia record is "CCTV5", then the target multimedia file is the live video of CCTV5.

In the method for identifying live multimedia files provided in this embodiment, a feature database is preset to store feature samples and identification information of live multimedia files, and the feature database is updated and maintained in real time by obtaining the real-time data stream of the live source background, when When a target multimedia file needs to be identified, the identification information of the corresponding live multimedia file is found in the feature database according to the characteristic information of the target multimedia file, so as to achieve the purpose of identifying the live multimedia file. To sum up, using the method for identifying live multimedia files provided by this embodiment, live multimedia files can be identified in real time, so that live video can also be identified in real time.

Embodiment two

This embodiment two provides a preferred embodiment of a method for identifying live multimedia files, which is a further preferred embodiment on the basis of embodiment one, and the specific improvements are:

First of all, in this embodiment, the fingerprint information of the audio data of the multimedia file is used as the feature information of the multimedia file, that is, the feature samples stored in the feature database are the fingerprint information of the audio data, and the feature information of the acquired target multimedia file is also Correspondingly, it is the fingerprint information of the audio data of the target multimedia file, so no matter when updating the feature database or when obtaining the feature information of the target multimedia file, it is necessary to obtain the audio data of the multimedia file and extract the fingerprint information of the audio data, which greatly reduces the The data transmission volume of the system is reduced, the traffic consumption is reduced, and the usability of the identification method is increased. Specifically, the audio data of the current period of the live multimedia file is divided into multiple audio segments, and the fingerprint information of each audio segment is extracted, so that the feature information of the current period of the live multimedia file is composed of the fingerprint information of the multiple audio segments.

Further, the time length (that is, the time length of n audio clips) of the feature sample in this embodiment is greater than the time length (that is, the time length of m audio clips) of the feature information of the acquired live multimedia file. A feature database (when the feature sample stored in the feature database is fingerprint information, the feature database is also called the fingerprint database, and the feature sample is called the fingerprint sample), the fingerprint information corresponding to the live multimedia file is added to the fingerprint sample, and The earliest fingerprint information in the fingerprint sample whose time length is the time length of adding characteristic information is deleted, thus ensuring the length of the fingerprint sample on the one hand, and ensuring the effective identification of the target multimedia file live at any time when matching the fingerprint sample. On the other hand, the real-time updating of fingerprint samples is guaranteed to ensure the real-time performance of identification.

Furthermore, when the fingerprint database is updated, a backup table combined with a timer is used to update, on the one hand, data conflicts can be avoided, and on the other hand, the update cycle can be controlled according to actual needs.

Specifically, FIG. 2 is a flow chart of a method according to Embodiment 2 of the present invention. As shown in FIG. 2 , the method specifically includes the following steps S202 to S216.

Step S202: Obtain the real-time data stream and identification information of the live multimedia file.

Step S204: Obtain the audio data of the current period of the live multimedia file according to the real-time data stream.

When the cloud server obtains the real-time data stream of the live multimedia file, it can further obtain the audio data of the current period of the live multimedia file. audio data, so that the obtained audio data is the data of the third predetermined time.

Preferably, when the audio data is acquired, format conversion can be performed on the audio data, and the acquired audio data can be converted into data in a unified format to facilitate subsequent processing; the audio data can also be denoised, for example, by using a sliding window The denoising technology removes the "spikes" in the audio data; it can also down-sample the audio data, which can reduce the amount of data storage and calculation under the premise of ensuring the accuracy of the data.

Step S206: Divide the audio data of the current period into m audio segments in sequence of time.

After the audio data of the third predetermined time is obtained, the audio segmentation module is called to divide the audio stream into m audio segments whose time length is the second predetermined time t in sequence.

It should be noted that, when executing steps S204 and S206, the following method can also be adopted: first divide the real-time data flow case in the current period into m data segments in time order, and then obtain the audio data of each data segment to obtain the audio fragment. This manner is equivalent to the above step S204 and step S206, and both are within the protection scope of the present application.

Step S208: extract the fingerprint information of each audio segment to obtain the feature information of the current period of the live multimedia file.

After obtaining m audio segments with a time length of t, extract the fingerprint information of each audio segment, wherein the fingerprint information of all audio segments constitutes the feature information of the current period of the live multimedia file, so that the feature information of the current period of the live multimedia file includes m fingerprint information, and the time length of the feature information is corresponding to the third predetermined time.

The characteristic information of the current period of the live multimedia file is composed of m fingerprint information arranged in chronological order, the first fingerprint information is the earliest fingerprint information, and the mth fingerprint information is the latest fingerprint information.

Wherein, the audio segment is preferably stereo data, and at the same time, the feature information of the target multimedia file is also the fingerprint information of the stereo data, and the unification of the two data sources can improve the matching accuracy.

When extracting the fingerprint information of an audio clip, the time domain feature of the audio can be extracted, for example, the amplitude of the audio clip can be extracted as the fingerprint information, and the time-frequency feature of the audio can also be extracted. The data processing speed of the former is fast, and the anti-noise ability of the latter is strong .

Step S210: Locate the multimedia record to be updated in the fingerprint database according to the identification information of the live multimedia file.

Wherein, the fingerprint database stores a plurality of multimedia records, and each multimedia record includes two parts: a feature sample of a multimedia file and identification information of a multimedia file corresponding to the feature sample, wherein the feature sample consists of n fingerprints arranged in chronological order The first fingerprint information in the feature sample is the earliest fingerprint information, and the last fingerprint information is the latest fingerprint information. The time length of each fingerprint information is t, the time length of the n fingerprint information is the first predetermined time T, and m<n, or the third predetermined time is less than the first predetermined time, that is, the characteristics of the current period of the live multimedia file The time length of the information is less than the time length of the fingerprint sample.

After the identification information is obtained, the multimedia record including the identification information can be located in the fingerprint database.

Step S212: Delete the earliest m fingerprint information of the fingerprint samples in the multimedia record to be updated.

When the multimedia record to be updated is located, the fingerprint sample in the multimedia record is updated. When updating, the first m fingerprint information in the fingerprint sample is deleted, that is, the current earliest m fingerprint information in the fingerprint sample is deleted. delete.

Step S214: Put the m fingerprint information of the current period of the live multimedia file into the fingerprint samples in the multimedia record to be updated in chronological order.

When updating, m pieces of fingerprint information in the current period of the live multimedia file are added to the end of the fingerprint sample, so that the added fingerprint information is the latest fingerprint information in the fingerprint sample.

It should be noted that, in this embodiment, step S212 may be executed first, and then step S214 may be executed, or step S214 may be executed first, and then step S212 may be executed. Wherein, when implementing step S214 and step S212, the following specific method steps can be adopted to realize:

Step S1: The feature pointer points to the first fingerprint information in the feature information of the current period of the live multimedia file, and the timer is cleared to start feature extraction timing;

Step S2: Obtain the fingerprint information pointed to by the feature pointer;

Step S3: Extracting feature samples of multimedia records corresponding to the identification information of the live multimedia to obtain a first feature sample;

Step S4: splicing the fingerprint information pointed to by the feature pointer to the end of the first feature sample to obtain a second feature sample;

Step S5: delete a fingerprint information from the beginning of the second feature sample;

Step S6: Determine whether the time in the timer reaches the third predetermined time, if it does not reach the third predetermined time, the feature pointer points to the next fingerprint information, and repeat steps S2 to S6; if it reaches the third predetermined time, use the obtained The second characteristic sample replaces the characteristic sample corresponding to the multimedia identifier in the multimedia record, wherein the third predetermined time is the playing time of the multimedia file corresponding to the m fingerprint information.

In the cloud server, the data streams and identification information of different live multimedia files are received in real time, and the fingerprint database is updated in time through steps S202 to S214, so that for any live multimedia files, the latest period of time is always stored in the fingerprint database. Fingerprint information of live multimedia files.

Step S216: Receive an identification request for identifying the target multimedia file, and identify the target multimedia file.

Specifically, the step S216 includes the steps S110 to S116 in the above-mentioned embodiment, which will not be repeated here.

Embodiment Three

This embodiment three provides an embodiment of a method for identifying live multimedia files. This embodiment is a further preferred embodiment on the basis of embodiment two. The specific improvements are:

First of all, the audio segment of the live multimedia file is formed by merging the left channel data and the right channel data. Correspondingly, the characteristic information of the target multimedia file is also the fingerprint information of the stereo data, and after merging the left and right channel data is For stereo data, set the weight parameter to adjust the proportion of the left and right channel data in the stereo data according to actual needs.

Further, when constructing the feature information of the target multimedia file, by setting multiple sets of weight data, the left and right channel data of the target multimedia file is converted into multiple sets of stereo data, and the fingerprint information corresponding to each set of stereo data is extracted, so that the target multimedia file The feature information includes multiple sets of fingerprint information. When identifying the target multimedia file, match each group of fingerprint information with the fingerprint samples in the fingerprint database, and use the multimedia record corresponding to the fingerprint sample with the maximum matching rate as the multimedia record corresponding to the target multimedia file to increase the accuracy of recognition .

Further, when extracting the fingerprint information of the stereo data of the audio segment, or when extracting the time-frequency feature data of each group of stereo data of the target multimedia file, the time-frequency feature of the stereo data is extracted, and according to the energy maximum point The moment, the frequency and energy of the location construct the fingerprint, so that the fingerprint can maintain good stability. And the constructed fingerprint is represented by hash code, which is convenient for data storage and processing.

Specifically, FIG. 3 is a flow chart of a method according to Embodiment 3 of the present invention. As shown in FIG. 3 , the method specifically includes the following steps S302 to S318.

Step S302: Obtain the real-time data stream and identification information of the live multimedia file, and obtain multiple audio segments of the current period of the live multimedia file according to the real-time data stream.

Specifically, the step S216 includes the steps S202 to S206 in the above-mentioned embodiment, which will not be repeated here.

Step S304: For each audio segment, combine the left channel data and the right channel data of the audio segment to obtain stereo data of the audio segment.

Specifically, the following formula can be used to obtain stereo data:

s=a*l+b*r, wherein, a+b=1, s is the stereo data of audio clip, and l is the left channel data of audio clip, and r is the right channel data of audio clip, and a and b are preset parameters.

Step S306: extracting the time-frequency feature data of the stereo data of each audio segment as the fingerprint information of the audio segment, so as to obtain the feature information of the current period of the live multimedia file.

When extracting the time-frequency feature data of the stereo data of the audio segment, the following steps are specifically included:

First, short-time Fourier transform is performed on the stereo data of the audio clip to obtain the time-frequency distribution map of the stereo data of the audio clip, and then the energy maximum points in the time-frequency distribution map are obtained, according to the maximum points at two different moments Value points A[ta,fa,Va], B[tb,fb,Vb] construct a fingerprint as fp[ta,fa,fb,tb-ta], and convert it into a hash code fp[hashData,ta], where , ta is the moment of the extreme point A, fa is the frequency of the extreme point A, Va is the energy of the extreme point A, tb is the moment of the extreme point B, fb is the extreme The frequency of point B with a large value, Vb is the energy of point B with a large extreme value, ta<tb, point A with a large maximum value and point B with a large extreme value are any two adjacent maximum energy values in the time-frequency distribution diagram Finally, combine all the constructed fingerprints in chronological order to obtain the fingerprint information of the audio clip.

Step S308: Update the fingerprint database according to the feature information of the current period of the live multimedia file.

Specifically, the step S216 includes the steps S210 to S214 in the above-mentioned embodiment, which will not be repeated here.

Step S310: Receive an identification request for identifying a target multimedia file.

Step S312: Obtain feature information of the target multimedia file according to the identification request.

Wherein, the feature information of the target multimedia file is the time-frequency feature data of the stereo data of the target multimedia file, and the method for specifically obtaining the video feature data is the same as the time-frequency feature data method of extracting the stereo data of the audio segment in step S306, and is no longer described here. repeat,

Wherein, the stereo data of the target multimedia file is formed by merging the left channel data and the right channel data in the audio data of the target multimedia file. Specifically, multiple sets of parameters can be used to obtain a plurality of stereo data, and correspondingly, the target multimedia file is obtained. The feature information is a plurality of fingerprint information.

When constructing the fingerprint information of the target multimedia file, the feature information of the target multimedia file is N fingerprint information, and one fingerprint information in the N fingerprint information is the time-frequency feature data of one stereo data in the N stereo data of the target multimedia file , wherein, the i-th stereo data among the N stereo data is si′=ai′*l′+bi′*r′, ai′+bi′=1, i=1, 2, 3...N.

Step S314: Match each fingerprint information of the target multimedia file with the fingerprint samples in the fingerprint database to obtain the matching rate of each fingerprint information.

Match the fingerprint information corresponding to each set of stereo data with the fingerprint samples in the fingerprint database. For any set of fingerprint information of stereo data, a matching fingerprint sample will be obtained, and each matched fingerprint sample will correspond to A matching rate, the multimedia record corresponding to the fingerprint sample with the highest matching rate (that is, the maximum matching rate) is used as the multimedia record corresponding to the target multimedia file.

Step S316: Use the multimedia record where the fingerprint sample corresponding to the maximum matching rate is located as the multimedia record corresponding to the target multimedia file.

Step S318: Obtain the identification information in the multimedia record corresponding to the target multimedia file to identify the target multimedia file.

Embodiment Four

The fourth embodiment provides a method for identifying a live multimedia file. In this method, the live multimedia file is a live video, and the time-frequency feature data of the audio data of the video is used as a fingerprint sample to construct the fingerprint database.

In this method, the audio data of the video all use stereo data to ensure that the audio source of the fingerprint information of the target video is consistent with the data format of the audio source of the fingerprint sample in the fingerprint database. At the same time, when preprocessing the audio data of the target video, for the impact of environmental noise on the quality of the audio data when the audio data of the target video is obtained by recording, the adaptive parameters are set to make the fingerprint information of the extracted target video more robust. Great.

To sum up, from two aspects of the user terminal and the fingerprint database, the fast and accurate identification of the video is realized. In this method, the real-time online identification of live webcast video is realized by updating the fingerprint database in real time.

Next, from the perspective of a system implementing the method of this embodiment, the method for identifying live multimedia files provided by this embodiment will be described in detail.

As shown in Figure 4, the system consists of four parts: user terminal, video search server, fingerprint management server, and video management server. Among them, the video search server, fingerprint management server, and video management server can jointly constitute a cloud server.

Specifically, the user terminal is responsible for acquiring audio data of the target video, and presenting video search results. The video search server is responsible for managing video identification requests from different user terminals, and sending these requests to the fingerprint management server; it is also used to receive the video identification results from the fingerprint management server, and return the results to the user terminal that made the identification request. On the one hand, the fingerprint management server is responsible for searching the fingerprint samples corresponding to the target video in the fingerprint database; on the other hand, it is responsible for creating, updating and maintaining the fingerprint database. On the one hand, the video management server is responsible for storing and managing the video data sent by the video source, storing the video data in the video database; at the same time, uploading the audio data and video information corresponding to the video to the fingerprint management server. The video search server and the fingerprint management server cooperate to realize the search of the target video, and the fingerprint management server and the video management server cooperate to realize the creation and update of the fingerprint database.

As shown in Figure 5, the user terminal includes the following modules:

Recording module: used to record audio data during video playback; audio preprocessing module: perform operations such as mixing, down-sampling, and noise reduction on the audio data acquired by the recording module to reduce the impact of noise in the recording environment on the matching results; fingerprint extraction module : Extract the fingerprint information of the preprocessed audio data; Result display module: Utilize hardware resources such as user terminal audio player, video player, display screen, etc. to display the results of video search (such as: play recognition results, present similar video); network transmission module: realize the data transmission requirement between the user terminal and the video search server, send the identification request of "scene information, target fingerprint (that is, the fingerprint information of the target video)" to the video search server, and receive the video Search the video recognition results sent by the server.

As shown in Figure 6, the video search server includes the following modules:

Network transmission module 1: used for information interaction with user terminals. Receive a video recognition request from a user terminal. Return the result of the video search to the user terminal; the video search management module: process the video recognition requests of a large number of users. Submit the video identification request sent by the user terminal to the fingerprint management server; identification result management module: process the video search result; network transmission module 2: used for information interaction with the fingerprint management server. Receive the video recognition result sent by the fingerprint management server. Send the video recognition request from the user terminal to the fingerprint management server.

As shown in Figure 7, the fingerprint management server includes the following modules:

Network transmission module 1: used for information interaction with the video search server. Receive a video recognition request. Return the result of the video search to the video search server; fingerprint search module: search for the target fingerprint in the fingerprint database, and return the identification result; fingerprint extraction module: extract the fingerprint information of the audio data from the video management server, and combine the generated fingerprint information with the video Information (that is, the identification information of the video) is transmitted to the fingerprint management module; the fingerprint management module: according to the required data of the fingerprint database generated by the video information and the fingerprint information, and stores the generated data in the fingerprint database; the network transmission module 2: Used for information exchange between the fingerprint management server and the video management server.

As shown in Figure 8, the video management server includes the following modules:

Network transmission module 1: realize the data transmission of video source and video management server; Video management module: according to the information (such as: channel, url etc.) of video source, video stream is stored in corresponding position in video database, simultaneously, The video stream and the video source information are transmitted to the audio extraction module; the audio extraction module: obtain the data from the video management server, extract the audio stream in the video stream, and transmit the audio stream and the video source information to the audio segmentation module; the audio preprocessing module : Mix the two-channel audio data into stereo, convert the audio data in different formats into a unified format, and down-sample the audio data, and pass the processed audio data together with the video source information to the audio segmentation module; the audio segmentation module : Divide or splice the audio data into audio segments with a time length of T in chronological order, upload the audio segments together with video source information to the fingerprint management server; network transmission module 2: realize data transmission between the video management server and the fingerprint management server.

When the method of this embodiment adopts the above-mentioned system to realize video recognition, it needs to go through the following steps: Step 1, obtain the audio data of the target video to be recognized; preprocess the obtained audio data; Step 3, obtain the audio data of the target video The fingerprint information of the audio data; Step 4, match the fingerprint information with the fingerprint samples in the fingerprint database constructed in advance (or updated in real time) to obtain the matching result; Step 5, return the matching result to the user terminal, the user The terminal can present and play relevant video content according to the obtained result.

The video data of the system is stored in multiple video databases in a distributed storage manner, and is managed by multiple video management servers. Different video databases share resources through a unified video list, and all video management servers share a video list. When the video list of one of the video databases is updated, the video management server corresponding to the video database broadcasts a list update message to the entire network (carrying the updated video list in the message), and other video databases update their own video lists according to the message .

The system has a unique fingerprint database, which is managed by the fingerprint management server. On the fingerprint management server, a fingerprint extraction module and a fingerprint search module are configured. The fingerprint extraction module is used to process the information sent by the video management server to form a fingerprint sample; the fingerprint search module is used to process the video recognition request sent by the video search server, search for the target fingerprint in the fingerprint database, and return the recognition result to the video search server .

Interaction between the video source and the video management server: the video source generates new video data, and submits a request message for uploading video data (including video source information and video content information) and video stream to the video management server; the video management server Extract the video source information, video content information and video stream in the message; update the video list of the video management server; on the one hand, store the above information and video stream in the video database, establish the association between the list information and the stored video data, The associated information is added to the local video list, and at the same time, information is updated to the entire network broadcast video list, and the form of the entire network video list is updated. On the other hand, the audio data in the video data is extracted, the above video source and video content information, the new video list information and the extracted audio data are packaged into a video library update message, and the message is submitted to the fingerprint management server through the network.

The interaction between the fingerprint management server and the video management server: (1) The fingerprint management server receives the video library update message, and generates the Track ID uniquely corresponding to the video according to the video source information and video content information in the message. Add the Track ID to the fingerprint list of the fingerprint management server. (2) Obtain the audio data in the message, and extract the fingerprint information of the audio data. For different types of video sources (live/non-live), different fingerprint information extraction schemes are adopted. (3) Encapsulate Track ID, fingerprint information, video source information and video content information, and store them in the fingerprint database. The fingerprint database is used to store the fingerprint information of the video data, the associated information of the video, and the related information of the video.

The fingerprint database is logically divided into two sub-databases (1) the live video fingerprint sub-database; (2) the video fingerprint sub-database. These two databases are jointly managed by the fingerprint management server to realize the operations of creation, update and maintenance.

The live video fingerprint sub-database stores the relevant information of the current live video: Track ID, fingerprint information, video information, and video related information. Each channel corresponds to a unique Track ID.

Track ID: The unique identification of the fingerprint information of the video in the fingerprint database.

Fingerprint information: only keep the fingerprint information of the latest live video with a length of T. The fingerprint information is updated accordingly with the update of the live video database. The specific implementation scheme is elaborated in the next part.

Video information: Information related to live video content and video storage information. Including: video channel, live broadcast name, live content, host, live channel url, storage location, etc.

Video-related information: links to other videos similar to the live video, information about products or locations appearing in the video, etc.

Live video fingerprint sub-database update:

(1) The video management server receives the video information and video stream from the live broadcast room (that is, the video source). Extract the video information part. Call the audio extraction module to extract the audio data of the video. Call the audio preprocessing module to preprocess the acquired audio data; call the video segmentation module to divide the video stream into audio segments with a time length of t (t is much smaller than T) in time order; add all audio segments to the network transmission In the sending queue of module 1; the video information and the audio clips of length t in the sending queue are sequentially encapsulated into a video library update message and uploaded to the fingerprint management server.

(2) The fingerprint management server receives the video storehouse update message, parses out the channel of the live video from the video information part in the message, and generates the corresponding Track ID; calls the fingerprint extraction module, and extracts the fingerprint information whose time length is t audio data ; The update period of the fingerprint database is P (T>>P=kt, k is an integer). Add the new fingerprint with a length of kt to the fingerprint list of the fingerprint database, and remove the original fingerprint information with a length of kt to ensure that the length of the fingerprint information in the fingerprint list is always T.

In this system, for the user terminal, by combining the fingerprint identification algorithm and the real-time update technology of the fingerprint database, the system can quickly identify the live video. At the same time, when the fingerprint database is created, the fingerprint information is extracted from the stereo data, so that the fingerprint information of the target video obtained by the user terminal is more likely to match its corresponding real fingerprint information in the fingerprint database, and the anti-noise ability of the identification process is improved.

Embodiment five

This embodiment five provides an embodiment of an identification device for a live multimedia file, the device can be set on a cloud server, as shown in Figure 9, the device includes an acquisition module 610, a positioning module 620, an update module 630, and a matching module 640 and identification module 650 .

Wherein, the cloud server communicates with the background database server of the live broadcast source to obtain the real-time data stream of the live multimedia file in real time, and can also obtain the identification information of the live multimedia file while obtaining the real-time data stream. The acquiring module 610 is configured to acquire feature information of the current period of the live multimedia file according to the input real-time data stream of the live multimedia file. For example, the video stream data of the live video is obtained, and the feature extraction module is used to process the video stream data to extract the corresponding audio fingerprint of the live video, and obtain the feature information of the current period of the live video.

A feature database is set in the cloud server, and the feature database is used to store at least one multimedia record. The multimedia record includes a feature sample of a multimedia file and identification information corresponding to the feature sample. The time length of the feature sample is a first predetermined time. The locating module 620 is used for locating the multimedia record to be updated in the feature database according to the identification information of the live multimedia file.

The update module 630 is used to update the feature samples in the multimedia record to be updated according to the feature information of the current period of the live multimedia file, so that for any live multimedia file, the feature information of the latest period of time is always stored in the feature database.

The matching module 640 is configured to receive an identification request for identifying a target multimedia file, and match the feature information of the target multimedia file included in the identification request with feature samples in the feature database to locate the multimedia record corresponding to the target multimedia file. The identifying module 650 is configured to acquire identification information of a multimedia file corresponding to the target multimedia file.

Adopt the identification device of the live multimedia file that this embodiment provides, preset a feature database to store the feature sample and the identification information of the live multimedia file, and update and maintain the feature database in real time by obtaining the real-time data stream of the live source background, when there is When the target multimedia file needs to be identified, the identification information of the corresponding live multimedia file is found in the feature database according to the characteristic information of the target multimedia file, so as to achieve the purpose of identifying the live multimedia file. To sum up, the apparatus for identifying live multimedia files provided by this embodiment can identify live multimedia files in real time, thereby also identifying live video in real time.

Preferably, the feature information is fingerprint information of audio data of the multimedia file, and the acquisition module 610 includes an audio data acquisition module, an audio segment segmentation module and a fingerprint information extraction module. Wherein, the audio data acquisition module is used to obtain the audio data of the current period of the live multimedia file according to the real-time data stream; the audio segment segmentation module is used to divide the audio data of the current period into a plurality of audio segments of the second predetermined time in chronological order; The fingerprint information extraction module is used to extract the fingerprint information of each audio segment to obtain the feature information of the current period of the live multimedia, wherein the second predetermined time is shorter than the first predetermined time.

Preferably, the feature sample is the fingerprint information of n audio clips, the feature information of the current period of the live multimedia file is the fingerprint information of m audio clips, m<n, the time length of n audio clips is the first predetermined time, update Module 630 includes a delete module and an add module. Wherein, the deletion module is used to delete the earliest m fingerprint information of the feature sample in the multimedia record to be updated; the addition module is used to place the m fingerprint information of the current period of the live multimedia file in chronological order in the multimedia record to be updated feature samples.

Further preferably, the update module 630 specifically performs the following steps:

Step S1: The feature pointer points to the first fingerprint information in the feature information of the current period of the live multimedia file, and the timer is cleared to start feature extraction timing;

Step S2: Obtain the fingerprint information pointed to by the feature pointer;

Step S3: Extracting feature samples of multimedia records corresponding to the identification information of the live multimedia to obtain a first feature sample;

Step S4: splicing the fingerprint information pointed to by the feature pointer to the end of the first feature sample to obtain a second feature sample;

Step S5: delete a fingerprint information from the beginning of the second feature sample;

Step S6: Determine whether the time in the timer reaches the third predetermined time, if it does not reach the third predetermined time, the feature pointer points to the next fingerprint information, and repeat steps S2 to S6; if it reaches the third predetermined time, use the obtained The second characteristic sample replaces the characteristic sample corresponding to the multimedia identifier in the multimedia record, wherein the third predetermined time is the playing time of the multimedia file corresponding to the m fingerprint information.

Further preferably, the fingerprint information extraction module includes a stereo data synthesis module and a time-frequency feature extraction module. Wherein, the stereo data synthesis module is used to merge the left channel data and the right channel data of the audio clip to obtain the stereo data of the audio clip; the time-frequency feature extraction module is used to extract the time-frequency feature data of the stereo data of the audio clip as audio The fingerprint information of the fragment.

Preferably, the feature information of the target multimedia file included in the identification request is N fingerprint information of the current period of the live multimedia file, and one of the N fingerprint information is one of the N stereo data of the target multimedia. Time-frequency feature data, wherein, the i-th stereo data in the N stereo data is si'=ai'*l'+bi'*r', ai'+bi'=1, and l' is the current of the live multimedia file The left channel data of the period, r' is the right channel data of the current period of the live multimedia file, ai' and bi' are preset parameters, i=1,2,3...N, the matching module 640 includes matching rate determination Modules and multimedia records identify modules. Among them, the matching rate determination module is used to match each fingerprint information of the target multimedia file with the feature samples in the feature database to obtain the matching rate of each fingerprint information; the multimedia record determination module is used to match the feature corresponding to the maximum matching rate The multimedia record where the sample is located serves as the multimedia record corresponding to the target multimedia file.

The above is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any person familiar with the technology can easily think of changes or replacements within the technical scope disclosed in the present invention. , should be covered within the protection scope of the present invention. Therefore, the protection scope of the present invention should be determined by the protection scope of the claims.

Claims (10)

1. A kind of 1. recognition methods of live multimedia file, it is characterised in that including：

   The feature of the live multimedia file present period is obtained according to the real-time stream of the live multimedia file of input Information；

   Multimedia recording to be updated is positioned in property data base according to the identification information of the live multimedia file, its In, the property data base is used to store at least one multimedia recording, and the multimedia recording includes multimedia file Feature samples, identification information corresponding with the feature samples, the time span of the feature samples was first scheduled time；

   Spy in the multimedia recording to be updated is updated according to the characteristic information of the live multimedia file present period Levy sample；

   The identification request of identification destination multimedia file is received, matches the destination multimedia text that the identification request includes The characteristic information of part and the feature samples in the property data base, to position the corresponding multimedia of the destination multimedia file Record；

   Obtain the identification information of the corresponding multimedia file of the destination multimedia file.
2. 2. the recognition methods of the live multimedia file according to claim 1, it is characterised in that the characteristic information is The finger print information of the voice data of multimedia file, the real-time stream of the live multimedia file according to input obtain institute The characteristic information of live multimedia file present period is stated, including：

   The voice data of the present period of the live multimedia file is obtained according to the real-time stream；

   The voice data of the present period is divided into multiple audio fragments of second scheduled time sequentially in time, its In, second scheduled time is less than first scheduled time；And

   The finger print information of each audio fragment of extraction, to obtain the characteristic information of the present period of the live multimedia.
3. 3. the recognition methods of the live multimedia file according to claim 2, it is characterised in that the feature samples are n The finger print information of a audio fragment, the characteristic information of the present period of the live multimedia file are the finger of m audio fragment Line information, m<N, the total time length of the n audio fragment is first scheduled time, according to the live multimedia The feature samples that the characteristic information of file is updated in the multimedia recording to be updated include：

   Delete the m earliest finger print information of feature samples in the multimedia recording to be updated；

   M finger print information of the present period of the live multimedia file is placed in chronological order described to be updated more In the feature samples of media recording.
4. 4. the recognition methods of the live multimedia file according to claim 3, it is characterised in that described according to described straight The characteristic information of playing multimedia file present period updates the feature samples in the multimedia recording to be updated, specific bag Include：

   Step S1：Feature pointer is directed toward first fingerprint letter in the characteristic information of the live multimedia file present period Breath, and timer is reset and starts feature extraction timing；

   Step S2：Obtain the finger print information that the feature pointer is directed toward；

   Step S3：Extraction and the feature samples of the corresponding multimedia recording of identification information of the live multimedia, to obtain Fisrt feature sample；

   Step S4：The finger print information that the feature pointer is directed toward is spliced to the end of the fisrt feature sample, to obtain Two feature samples；

   Step S5：A finger print information is deleted from the starting of the second feature sample；

   Step S6：Judge whether the time in timer reached for the 3rd scheduled time, if not up to described 3rd scheduled time, The feature pointer is directed toward next finger print information, and repeats step S2 to S6；If reaching the 3rd scheduled time, The corresponding feature samples of multi-media tag described in the multimedia recording are replaced with the obtained second feature sample, its In, the 3rd scheduled time is the total duration of the reproduction time of the corresponding multimedia file of the m finger print information.
5. 5. the recognition methods of the live multimedia file according to claim 2, it is characterised in that extract the audio piece The finger print information of section includes：

   Merge the left channel data and right data of the audio fragment, to obtain the stereo data of the audio fragment； And

   Extract finger print information of the time-frequency characteristics data of the stereo data of the audio fragment as the audio fragment.
6. 6. the recognition methods of the live multimedia file according to claim 2, it is characterised in that in the identification request Including the destination multimedia file characteristic information for the live multimedia file present period N fingerprint believe Cease, a finger print information in the N finger print information is one in N stereo data of the destination multimedia The time-frequency characteristics data of stereo data, wherein, i-th stereo data in the N stereo data for si '= Ai ' * l '+bi ' * r ', wherein, ai '+bi '=1, l ' they are the L channel number of the present period of the live multimedia file According to the right data for the present period that, r ' is the live multimedia file, ai ' and bi ' are default parameter, i=1, 2,3… N,

   In the method, the characteristic information for the destination multimedia file that the matching identification request includes and institute The feature samples in property data base are stated, are included with positioning the corresponding multimedia recording of the destination multimedia file：

   Each finger print information of the destination multimedia file is matched with the feature samples in the property data base respectively, is obtained To the matching rate of each finger print information；

   Multimedia recording where the corresponding feature samples of maximum matching rate is corresponding more as the destination multimedia file Media recording.
7. A kind of 7. identification device of live multimedia file, it is characterised in that including：

   Acquisition module, the real-time stream for the live multimedia file according to input obtain the live multimedia file and work as The characteristic information of preceding period；

   Locating module is to be updated more for being positioned according to the identification information of the live multimedia file in property data base Media recording, wherein, the property data base is used to store at least one multimedia recording, and the multimedia recording includes The feature samples of multimedia file, identification information corresponding with the feature samples, the time spans of the feature samples are the One scheduled time；

   Update module, for updating more matchmakers to be updated according to the characteristic information of the live multimedia file present period Feature samples in body record；

   Matching module, for receiving the identification request of identification destination multimedia file, matches the institute that the identification request includes The characteristic information of destination multimedia file and the feature samples in the property data base are stated, to position the destination multimedia text The corresponding multimedia recording of part；

   Identification module, for obtaining the identification information of the corresponding multimedia file of the destination multimedia file.
8. 8. the identification device of the live multimedia file according to claim 7, it is characterised in that the characteristic information is The finger print information of the voice data of multimedia file, the acquisition module include：

   Voice data acquisition module, for the present period according to the real-time stream acquisition live multimedia file Voice data；

   Audio fragment splits module, for the voice data of the present period to be divided into the second pre- timing sequentially in time Between multiple audio fragments, wherein, second scheduled time is less than first scheduled time；And

   Finger print information extraction module, for extracting the finger print information of each audio fragment, to obtain the live multimedia Present period characteristic information.
9. 9. the identification device of the live multimedia file according to claim 8, it is characterised in that the feature samples are n The finger print information of a audio fragment, the characteristic information of the present period of the live multimedia file are the finger of m audio fragment Line information, m<N, the total time length of the n audio fragment is first scheduled time, and the update module includes：

   Removing module, for deleting the m earliest finger print information of feature samples in the multimedia recording to be updated；

   Add module, for m finger print information of the present period of the live multimedia file to be placed in institute in chronological order In the feature samples for stating multimedia recording to be updated.
10. 10. the identification device of the live multimedia file according to claim 9, it is characterised in that the update module tool Body performs following steps：

    Step S1：Feature pointer is directed toward first fingerprint letter in the characteristic information of the live multimedia file present period Breath, and timer is reset and starts feature extraction timing；

    Step S2：Obtain the finger print information that the feature pointer is directed toward；

    Step S3：Extraction and the feature samples of the corresponding multimedia recording of identification information of the live multimedia, to obtain Fisrt feature sample；

    Step S4：The finger print information that the feature pointer is directed toward is spliced to the end of the fisrt feature sample, to obtain Two feature samples；

    Step S5：A finger print information is deleted from the starting of the second feature sample；

    Step S6：Judge whether the time in timer reached for the 3rd scheduled time, if not up to described 3rd scheduled time, The feature pointer is directed toward next finger print information, and repeats step S2 to S6；If reaching the 3rd scheduled time, The corresponding feature samples of multi-media tag described in the multimedia recording are replaced with the obtained second feature sample, its In, the 3rd scheduled time is the total duration of the reproduction time of the corresponding multimedia file of the m finger print information.