WO2019227579A1 - Conference information recording method and apparatus, computer device, and storage medium - Google Patents

Abstract

A conference information recording method and apparatus, a computer device, and a storage medium. The method comprises: if it is detected that a collecting channel generates voice data, obtaining the voice data and current time; extracting a voiceprint feature of the voice data; moreover, recognizing the voiceprint feature by obtaining a preset voiceprint recognition model to determine a target participant identifier corresponding to the voiceprint feature, so as to quickly and accurately confirm the identity of a current spokesman, thereby improving the efficiency and accuracy of conference information recording for recording the spokesman; and moreover, recognizing the voice data in real time to generate a real-time voice text, and associating the participant identifier, the current time, and the real-time voice text to obtain speaking information of the participant corresponding to the participant identifier at a current conference, thereby avoiding missing and wrong recording caused by manual recording, and improving the efficiency and accuracy of conference information recording.

Description

Conference information recording method, device, computer equipment and storage medium

This application is based on a Chinese invention patent application filed on May 29, 2018 with the application number 201810528516.1 and entitled "Meeting Information Recording Method, Device, Terminal Equipment and Storage Medium", and claims priority.

Technical field

The present application relates to the field of computer technology, and in particular, to a method, a device, a computer device, and a storage medium for recording conference information.

Background technique

With the advancement of science and technology and the rapid development of computer network technology, the demand for network computing technology for interconnecting computers is increasing. Under the network connection environment, computer resources sharing and information sharing, circulation, distribution, and exchange can be smoothly performed.

For example, a system for implementing multi-person collaborative work through a network is a remote conference system. A remote conference system is used to transfer images and sounds between conference rooms located at remote locations, and the participants' faces reflected on the TV screen can be used to hold meetings. Participants who are physically located remotely from each other can discuss matters as if they were meeting in the same conference room. In addition, since each participant does not need to move, time and expenses such as travel expenses can be effectively reduced.

Remote conferences have brought us great convenience. However, when there are many participants in remote conferences, when recording conference information, they often misrepresent the speaker due to the large number of participants. When the speech interval is short or the speech time is long, it is easy to miss some important information, which makes the efficiency of meeting information recording low, and the accuracy rate cannot meet the requirements.

Summary of the Invention

Based on this, it is necessary to provide a meeting information recording method, device, computer equipment, and storage medium that can improve the efficiency and accuracy of meeting information recording in response to the above technical problems.

A conference information recording method includes:

If a voice data transmission request for a collection channel is received, obtaining the voice data and the current time;

Performing voiceprint feature extraction on the voice data to obtain a target voiceprint feature of the voice data;

According to a preset voiceprint recognition model, voiceprint recognition is performed on the target voiceprint feature to determine a target participant identification corresponding to the target voiceprint feature, wherein the voiceprint recognition model includes a basic voiceprint of the participant Characteristics, participant identification, and the corresponding relationship between the basic voiceprint characteristics and the participant identification;

Performing speech recognition on the speech data to generate real-time speech text;

The target participant identification, the current time, and the real-time voice text are associated and recorded to obtain the speech information of the target participant corresponding to the target participant identification.

A conference information recording device includes:

A data collection module, configured to obtain the voice data and the current time if a voice data transmission request of a collection channel is received;

A feature extraction module, configured to perform voiceprint feature extraction on the voice data to obtain a target voiceprint feature of the voice data;

A person determination module is configured to perform voiceprint recognition on the target voiceprint feature according to a preset voiceprint recognition model, and determine a target participant identification corresponding to the target voiceprint feature, where the voiceprint recognition model includes The basic voiceprint characteristics of the participants, the participant identification, and the correspondence between the basic voiceprint characteristics and the participant identification;

A voice recognition module, configured to perform voice recognition on the voice data to generate real-time voice text;

The record generating module is configured to record the target participant identification, the current time, and the real-time voice text in association with each other, and obtain the speech information of the target participant corresponding to the target participant identification.

A computer device includes a memory, a processor, and computer-readable instructions stored in the memory and executable on the processor, and the processor implements the method for recording conference information described above when the processor executes the computer-readable instructions A step of.

One or more non-volatile readable storage media storing computer-readable instructions that, when executed by one or more processors, cause the one or more processors to execute the conference information Document the steps of the method.

Details of one or more embodiments of the present application are set forth in the accompanying drawings and description below, and other features and advantages of the present application will become apparent from the description, the drawings, and the claims.

BRIEF DESCRIPTION OF THE DRAWINGS

In order to explain the technical solutions of the embodiments of the present application more clearly, the drawings used in the description of the embodiments of the application will be briefly introduced below. Obviously, the drawings in the following description are just some embodiments of the application. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without paying creative labor.

FIG. 1 is a schematic diagram of an application environment of a conference information recording method according to an embodiment of the present application; FIG.

FIG. 2 is an implementation flowchart of a conference information recording method according to an embodiment of the present application; FIG.

3 is a flowchart of implementing step S2 in the conference information recording method according to the embodiment of the present application;

4 is a flowchart of implementing step S4 in a conference information recording method according to an embodiment of the present application;

5 is an implementation flowchart of constructing a preset voiceprint recognition model in a conference information recording method according to an embodiment of the present application;

FIG. 6 is a flowchart of real-time recording correction in a conference information recording method according to an embodiment of the present application; FIG.

7 is a schematic diagram of a conference information recording device according to an embodiment of the present application;

FIG. 8 is a schematic diagram of a computer device according to an embodiment of the present application.

Detailed ways

In the following, the technical solutions in the embodiments of the present application will be clearly and completely described with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are part of the embodiments of the present application, but not all of the embodiments. Based on the embodiments in this application, all other embodiments obtained by a person of ordinary skill in the art without creative efforts shall fall within the protection scope of this application.

Please refer to FIG. 1, which illustrates an application environment of a conference information recording method according to an embodiment of the present application. The conference information recording method is applied to a conference information recording scenario of a remote conference. The recording scenario includes the server, client, and monitor. Among them, the server and the client, and the server and the monitor are connected through the network. Participants send voice data to the server through the client, and the server passes Analyze the voice data, and convert the voice data into the corresponding meeting record information. The meeting information recorder monitors and modifies the server's meeting record information through the monitoring terminal. The client can specifically but not limited to various microphones, microphones, and mobile phones. For smart devices with recording functions, the server and monitor can be implemented with independent servers or a server cluster consisting of multiple servers.

Please refer to FIG. 2, which illustrates a conference information recording method provided by an embodiment of the present application. The method is applied to the server in FIG. 1 as an example for description, as follows:

S1: If a voice data transmission request is received from the acquisition channel, the voice data and the current time are obtained.

Specifically, when the collection channel sends a data transmission request to the server, the server receives the voice data in the request and obtains the current time, where the collection channel is the client of this embodiment.

The collection channel collects conference speech information of the participants, which may specifically be various microphones, microphones, or other communication equipment with a voice transmission function, etc., which are not specifically limited here.

It is worth noting that in the embodiment of the present application, multiple collection channels are used to collect voice data. In general, there are a large number of remote conferences. If more than 4 participants use the same collection channel, the participants who are far away from the collection channel will make the participants farther from the collection channel. The voice data is easily recognized as the background sound, so it is filtered out when the voice signal is optimized, so that the accuracy of voice recognition becomes low. In the embodiment of the present application, multiple channels are used to avoid participants who use the same acquisition channel. The problem of excessive voice confusion recognition rate reduction, the collection channel can be set up for each participant, an independent collection channel, or two to four people in the vicinity use the same collection channel, which can be set according to the actual situation. There are no restrictions here.

S2: Perform voiceprint feature extraction on the voice data to obtain the target voiceprint feature of the voice data.

Specifically, voiceprint feature extraction is performed on the voice data, and the specific process is: by performing voice encoding and preprocessing of the voice data on the voice data, the voice data is converted from an analog signal into a digital signal, and then the voiceprint is performed from the digital signal. Feature extraction, and then get the target voiceprint feature of speech data.

The voiceprint features include, but are not limited to, acoustic features, morphological features, prosodic features, dialect accent information, and channel information.

Preferably, the voiceprint feature used in this application is an acoustic feature.

S3: According to a preset voiceprint recognition model, voiceprint recognition is performed on the target voiceprint feature to determine the target participant identification corresponding to the target voiceprint feature. The voiceprint recognition model includes the basic voiceprint features of the participants. Personnel identification, and the corresponding relationship between the basic voiceprint characteristics and participant identification.

Specifically, the preset voiceprint recognition model includes: the basic voiceprint characteristics of the participant, the participant identification, and the correspondence between the basic voiceprint characteristics and the participant identification. After the target voiceprint feature is obtained, the target voiceprint feature is subjected to voiceprint recognition through a preset voiceprint recognition model, thereby obtaining the target participant identification corresponding to the target voiceprint feature, and then determining the current speaker. Meeting with personnel.

Among them, voiceprint recognition, a type of biometric technology, also known as speaker recognition, has two types, namely speaker recognition and speaker confirmation. Voiceprint recognition is to convert acoustic signals into electrical signals, and then use a computer to identify them.

It is easy to understand that the voiceprint recognition is used in the embodiment of the present application to identify the participant who is currently speaking, that is, speaker identification.

S4: Perform speech recognition on speech data to generate real-time speech text.

Specifically, by performing voice recognition on voice data, real-time voice texts of speech information of participants are generated, thereby automatically recording speech information and improving the efficiency of meeting records.

Among them, the real-time speech text refers to segmenting the speech data of the conference speech according to the pause and silence during each conference speech, and the segmented segmented speech data is obtained, and each segmented speech data is subjected to speech recognition to obtain The corresponding recognition text is speech recognition text.

For example, in a specific embodiment, a piece of voice data is acquired from 0 to 1.8 seconds, and is recorded as voice data E. The voice data acquired from 1.8 to 3 seconds is empty, and 3 to 8 seconds are obtained. The voice data at the other end is obtained, recorded as voice data F, and the voice data E is subjected to voice recognition, and a real-time voice text is obtained: “in this quarter”, and the voice data F is subjected to voice recognition, and a real-time voice text is obtained: Our business goals have been exceeded. "

The voice data may be recognized by a voice recognition algorithm or a third-party tool with a voice recognition function, which is not limited in particular. Speech recognition algorithms include, but are not limited to, speech recognition algorithms based on channel models, speech template matching recognition algorithms, or artificial neural network speech recognition algorithms.

Preferably, the speech recognition algorithm used in the embodiment of the present application is a speech recognition algorithm based on a channel model.

S5: The target participant identification, the current time, and the real-time voice text are associated and recorded to obtain the speech information of the target participant corresponding to the target participant identification.

Specifically, for each real-time voice text, the real-time voice text and the current time corresponding to the real-time voice text are associated with the target participant ID, and the speech information of the target participant corresponding to the target participant ID is obtained. .

For example, in a specific embodiment, the obtained real-time voice text is "Now I announce the official start of the meeting", the corresponding current time is "April 18, 2018 13:52:00", and the corresponding participant ID is "Work ID 003", which records the real-time voice text and its corresponding current time and participant identification, and obtains a record of the meeting information: "April 18, 2018 13:52:00, Work ID 003, now I Announce the official start of the meeting. "

Preferably, after the meeting is ended, all meeting information records can be sorted in chronological order.

In this embodiment, on the one hand, if it is detected that the acquisition channel generates voice data, the voice data and the current time are obtained, and the voiceprint feature extraction is performed on the voice data to obtain the target voiceprint feature of the voice data. A preset voiceprint recognition model is used to identify the target voiceprint feature, so as to determine the target participant identification corresponding to the target voiceprint feature, so that when there are many participants in a remote conference, voiceprint recognition is used to quickly Accurately identify the current speaker's identity, improve the efficiency of identifying and recording the speaker in the conference information record, and improve the accuracy of the conference information record. On the other hand, the real-time recognition of voice data and the generation of real-time voice text will be provided The identification, current time, and real-time voice and text are associated to obtain the speech information of the participant at the conference corresponding to the participant identification, avoiding missed and misremembered records caused by manual recording, and improving the efficiency and accuracy of conference information recording.

In an embodiment, as shown in FIG. 3, in step S2, voiceprint feature extraction is performed on the voice data to obtain the target voiceprint feature of the voice data, which specifically includes the following steps:

S21: Perform voiceprint analysis on the voice data to obtain the initial voiceprint.

Specifically, the voice data obtained by the device is an analog signal, and voiceprint analysis is performed on the voice data, that is, after the voice data is obtained, the voice data is encoded using Pulse Code Modulation (PCM) to make these The analog signal is converted into a digital signal, and the analog signal in the voice data is sampled at a sampling point every predetermined time to discretize it, and then the sampled signal is quantized, and the quantized digital is output in a binary code group manner. Signal, according to the frequency range of speech of 200-3400Hz, the sampling rate can be set to 8KHz, and the quantization accuracy is 16bit.

It should be understood that the numerical ranges of the sampling rate and quantization accuracy herein are the preferred ranges of the present application, but can be set according to the needs of practical applications, and are not limited here.

S22: Perform pre-emphasis processing on the initial voiceprint to generate an emphasized voiceprint with a flat spectrum.

Specifically, since the glottic excitation and the nasal and nasal radiation will affect the average power spectrum of the speech signal, the high frequency will drop by 6dB / octave when the frequency exceeds 800Hz. Therefore, when calculating the speech signal spectrum, the higher the frequency the corresponding The smaller the component, the pre-emphasis is performed in the pre-processing for this purpose. The purpose of pre-emphasis is to improve the high-frequency part, make the signal spectrum flat, and keep it in the entire low-frequency to high-frequency band. Can use the same signal-to-noise ratio to find the spectrum to facilitate spectrum analysis or channel parameter analysis. Pre-emphasis can be performed before the anti-aliasing filter when the voice signal is digitized. This not only can perform pre-emphasis, but also can compress the dynamic range of the signal and effectively improve the signal-to-noise ratio. Pre-emphasis can be implemented using a first-order digital filter, such as a Finite Impulse Response (FIR) filter.

It is worth noting that the voice signals obtained by the device are all analog signals. Before pre-emphasis processing of these analog signals, the analog information needs to be converted into digital signals through sampling and quantization. According to the speech spectrum range 200-3400Hz, sampling The rate can be set to 8KHz and the quantization accuracy is 16bit.

It should be understood that the numerical ranges of the sampling rate and quantization accuracy herein are the preferred ranges of the present application, but can be set according to the needs of practical applications, and are not limited here.

After the pre-emphasis of the speech signal, the high-frequency part of the frequency spectrum is improved, and the signal also becomes flat, generating a weighted processing voiceprint with a flat spectrum, which is beneficial for subsequent voiceprint feature extraction.

S23: Framing and windowing are used to frame and emphasize the processed voiceprint to obtain a first basic voice frame.

Specifically, the voiceprint signal has the property of short-term stability. After pre-emphasis processing, the voiceprint signal needs to be framed and windowed to maintain the short-term stability of the signal. Generally, every second The number of frames included is between 33 and 100 frames. In order to maintain the continuity between frames, so that the adjacent two frames can smoothly transition, the overlapping frame method is adopted, that is, to maintain a crossover between the adjacent two frames, which is the frame shift. .

Preferably, the range of the ratio of the frame shift to the frame length is (0, 0.5).

For example, in a specific implementation, the pre-emphasized voiceprint signal is s' (n), the frame length is N sampling points, and the frame shift is M sampling points. When the sampling point corresponding to the l frame is the nth, the corresponding relationship between the original speech signal x _l (n) and each parameter is:

x _l (n) = x [(l-1) M + n]

Among them, n = 0,1, ..., N-1, N = 256.

After the voiceprint is framed, the corresponding window function w (n) is multiplied with the pre-emphasized voiceprint signal s' (n) to obtain the windowed speech signal _Sw . This voiceprint signal is used as the first A basic speech frame.

The window functions include, but are not limited to, rectangular windows, Hamming windows, and Hanning windows.

The rectangular window expression is:

Among them, w (n) is a window function, N is the number of sampling points, and n is the nth sampling point.

The Hamming window expression is:

Wherein, pi is the perimeter, and preferably, the value of pi in the embodiment of the present application is 3.1416.

The Hanning window expression is:

The pre-emphasis processing of the voiceprint is framed and windowed, so that the voiceprint maintains the continuity between frames and eliminates some abnormal signal points, which improves the robustness of the voiceprint.

S24: Silent tone separation is performed on the first basic voice frame to obtain a first target voice frame.

Specifically, the silent sound separation is performed on the first basic speech frame to obtain the first target speech frame. For details about the implementation process of step S42 and step S43, to avoid repetition, details are not described herein again.

S25: Perform voiceprint feature extraction on each first target voice frame to obtain a target voiceprint feature.

Specifically, after pre-emphasis processing, framing, windowing, and silence separation, a stable voiceprint is obtained, and voiceprint features are extracted using the voiceprint.

Among them, voiceprint feature extraction is to extract and select acoustic or linguistic features with strong separability and high stability for the speaker's voiceprint.

Preferably, the voiceprint feature selected in this application is a linear cepstrum feature in the acoustic feature.

In this embodiment, the voiceprint analysis is performed on the speech data of the participants to obtain the initial voiceprint, and the initial voiceprint samples are pre-emphasized to generate a weighted voiceprint with a flat frequency spectrum. The windowing method performs frame processing on the weighted voiceprint to obtain a robust first basic speech frame, and then performs silent sound separation on the first basic speech frame, and filters out silent noises to obtain the first The target voice frame makes the voiceprint continuous and robust, and then extracts the target voiceprint characteristics of the participants from the voiceprint, improving the accuracy and robustness of the target voiceprint characteristics, which is conducive to subsequent voice In the pattern recognition, the accuracy of voiceprint recognition is improved to ensure the accurate recognition of target participants.

In an embodiment, as shown in FIG. 4, in step S4, the voice data is subjected to voice recognition to generate real-time voice text, which specifically includes the following steps:

S41: Perform speech analysis on the speech data to obtain a frame set including a second basic speech frame.

Specifically, speech analysis is performed on the acquired speech data to obtain a frame set including a second basic speech frame. The speech analysis includes, but is not limited to, speech encoding and pre-processing of speech signals.

Among them, speech encoding is to encode analog speech signals and convert analog signals into digital signals, thereby reducing the transmission code rate and performing digital transmission. The basic methods of speech encoding can be divided into waveform encoding, parametric encoding, and hybrid encoding.

Preferably, the speech coding method used in this proposal is waveform coding. Waveform coding is a digital voice signal formed by sampling, quantizing, and encoding the waveform signal of the analog voice in the time domain. The waveform coding can provide high voice quality.

Among them, the preprocessing of a voice signal refers to pre-emphasis, framing, windowing and other preprocessing operations on the voice signal before analysis and processing. The purpose of speech signal pre-processing is to eliminate the effects of aliasing, higher harmonic distortion, high frequency and other factors caused by the human vocal organ itself and the equipment that collects the speech signal, so as to make the follow-up The signal obtained by speech processing is more uniform and smooth, which provides high-quality parameters for signal parameter extraction and improves the quality of speech processing.

S42: Perform silence detection on the second basic voice frame to obtain K consecutive silence frames in the second basic voice frame, where K is a natural number.

Specifically, during the continuous speaking process of the participants, the voice signals in the voice data can be divided into two states: the active period and the silent period. The silent period does not transmit any voice signals. The uplink and downlink activation periods and silent periods are mutually related. independent. During the speech process, the participants will have a pause state. This state will bring about a pause in the voice signal, that is, a silent period. When performing speech recognition and converting text, the silent period state needs to be detected, and then the silent period and the active period will be detected. Separation is performed to obtain a continuous activation period, and the voice signal of the remaining continuous activation period is used as a target speech frame.

The methods for detecting the state of the silence include, but are not limited to, voice endpoint detection, detection of audio mute algorithms, and voice activity detection (VAD) algorithms.

Preferably, the specific implementation process of performing silence detection on the second basic voice frame and obtaining K consecutive silence frames in the second basic voice frame used in the embodiments of the present application includes steps A to E, which are detailed as follows:

Step A: Calculate the frame energy of the second basic speech frame of each frame.

Specifically, the frame energy is the short-term energy of the voice signal, and reflects the data amount of the voice information of the voice frame. The frame energy can be used to determine whether the voice frame is a sentence frame or a mute frame.

Step B: For each frame of the second basic voice frame, if the frame energy of the second basic voice frame is less than a preset frame energy threshold, mark the second basic voice frame as a mute frame.

Specifically, the frame energy threshold is a preset parameter. If the calculated frame energy of the second basic voice frame is less than a preset frame energy threshold, the corresponding second basic voice frame is marked as a mute frame, and the frame energy is The threshold may be specifically set according to actual requirements. For example, the frame energy threshold is set to 0.5, or the frame energy of each second basic voice frame may be calculated and set based on calculations. There is no limitation here.

For example, in a specific embodiment, the frame energy threshold is set to 0.5, and the frame energy calculation is calculated for 6 second basic speech frames: J ₁ , J ₂ , J ₃ , J ₄ , J _5, and J ₆ , and the results are obtained respectively. It is: 1.6, 0.2, 0.4, 1.7, 1.1, and 0.8. From this result, it is easy to understand that the second basic speech frame J ₂ and the second basic speech frame J ₃ are silent frames.

Step C: If H consecutive mute frames are detected, and the cut H is greater than a preset continuous threshold I, the frame set composed of the H consecutive mute frames is regarded as a continuous mute frame.

Specifically, the continuous threshold I can be preset according to actual needs. If the number of continuous silent frames is H, and the cut H is greater than the preset continuous threshold I, all of the intervals composed of the H continuous silent frames are all Mute frames are merged to get a continuous mute frame.

For example, in a specific embodiment, the preset continuous threshold I is 5, and at a certain moment, the status of the acquired mute frame is shown in Table 1. Table 1 shows a frame composed of 50 second basic voice frames. The set, as shown in Table 1, contains five or more consecutive mute frame intervals: the interval P composed of the second basic voice frame corresponding to frame number 7 to frame number 13 and frame number 21 to frame number 29 The interval Q composed of the corresponding second basic voice frame, therefore, the 7 second basic voice frames corresponding to the frame number 7 to the frame number 13 included in the interval P are combined to obtain a continuous mute frame P, and the continuous mute frame P The duration of P is the sum of the durations of the 7 second basic speech frames corresponding to the frame number 7 to frame number 13. According to this method, the second basic speech frame corresponding to the frame number 21 to frame number 29 included in the interval Q is performed. The combination is another continuous silence frame Q, and the duration of the continuous silence frame Q is the sum of the durations of the 9 second basic voice frames corresponding to the frame number 21 to the frame number 29.

Table I

Frame number 1 2 3 4 5 6 7 8 9 10 Whether to mute the frame no no Yes no no no Yes Yes Yes Yes Frame number 11 12 13 14 15 16 17 18 19 20 Whether to mute the frame Yes Yes Yes no no no no no no no Frame number twenty one twenty two twenty three twenty four 25 26 27 28 29 30 Whether to mute the frame Yes Yes Yes Yes Yes Yes Yes Yes Yes no Frame number 31 32 33 34 35 36 37 38 39 40 Whether to mute the frame Yes Yes no no no no no no Yes Yes Frame number 31 32 33 34 35 36 37 38 39 40 Whether to mute the frame no Yes Yes no no Yes no no no no

Step D: According to the method of steps A to C, obtain a total of K consecutive silent frames.

Taking Table 1 listed in step C as an example, the continuous mute frames obtained are continuous mute frame P and continuous mute frame Q, because in the example corresponding to step C, the value of K is 2.

S43: Divide the second basic voice frame included in the frame set into K + 1 second target voice frames according to the K silence frames.

Specifically, the K consecutive silent frames obtained in step S42 are used as the demarcation points, and the second basic voice frame included in the frame set is divided to obtain a set interval of K + 1 second basic voice frames. All the second basic speech frames included in the collection interval are used as a target speech frame.

For example, in a specific implementation manner, the status of the acquired mute frame is shown in Table 1 in step C in S42, which shows two consecutive mute frames, which are 7 corresponding to frame number 7 to frame number 13, respectively. The second basic voice frame is combined to obtain a continuous mute frame P, and the nine second basic voice frames corresponding to frame number 21 to frame number 29 are combined to obtain a continuous mute frame Q. The two consecutive mute frames are used as the demarcation point. The frame set containing 50 second basic speech frames is divided into three intervals, which are: the interval M ₁ composed of the second basic speech frames corresponding to the frame number 1 to frame number 6, and the frame number 14 to the frame number. The interval M ₂ composed of the second basic speech frame corresponding to 20 and the interval M ₃ composed of the second basic speech frames corresponding to the frame number 30 to frame number 50 combine all the second basic speech frames in the interval M ₁ , A combined speech frame is obtained as the second target speech frame M ₁ .

S44: Convert each second target voice frame into real-time voice text.

Specifically, text conversion is performed on each target voice frame to obtain a real-time voice text corresponding to the target voice frame.

Among them, the text conversion can use a tool that supports speech conversion text, or can be used for text conversion algorithms, which are not specifically limited here.

In this embodiment, the voice data is parsed to obtain a frame set including a second basic voice frame, and then the second basic voice frame is muted to detect K consecutive silent frames in the second basic voice frame. These K consecutive mute frames divide the second basic voice frame contained in the frame set into K + 1 second target voice frames, and convert each target voice frame into a real-time voice text, so that the received voice The signals are converted into independent real-time voice texts in real time, which enables accurate recording of the speech information of participants in real time, and improves the accuracy and efficiency of meeting information recording.

In an embodiment, as shown in FIG. 5, before step S3, the conference information recording method further includes the following steps:

S61: Collect voiceprint features for each participant to obtain the basic voiceprint features for each participant.

Specifically, before the meeting begins, a predetermined corpus is provided to the participants, and the participants are recorded and collected when the corpus is read aloud to obtain a recording sample, and then the voiceprint feature extraction of the voice signal in the recorded sample is obtained The basic voiceprint characteristics of the meeting staff.

The preset corpus can be an article, a web news, or a meeting minutes, etc., and can be set according to actual needs, and is not specifically limited here.

For example, in a specific embodiment, a professional paper is provided as a preset corpus. For each participant, a recording sample of the participant reading the preset corpus is collected, and the participant is obtained based on the recording sample. The basic voiceprint features.

The voiceprint feature extraction of the voice signal in the recorded text to obtain the basic voiceprint feature of the participant can refer to the description of steps S21 to S25 or the description of step S2. To avoid repetition, I won't repeat them here.

It is worth noting that the basic voiceprint feature is a collection containing multiple voiceprint features.

S62: Obtain the identity of each participant, and build an initial voiceprint recognition model based on the corresponding relationship between the participant's basic voiceprint characteristics, the participant's identification, and the participant's basic voiceprint characteristics and the participant's identification.

Specifically, for each participant, the participant identification of the participant is obtained, and based on the participant identification, the basic voiceprint characteristics of the participant, and the correspondence between the basic voiceprint characteristics of the participant and the participant identification Relationship, construct a map of the voiceprint feature information of the participants, and store the map in the initial voiceprint model.

S63: Use the genetic algorithm to train the initial voiceprint recognition model to generate a preset voiceprint recognition model.

Specifically, after the initial voiceprint model is obtained, the initial voiceprint model needs to be trained in order to improve the recognition rate of the voiceprint recognition model. The training methods include, but are not limited to, a template matching method, a nearest neighbor method, and a neural network method. , Hidden Markov Model (HMM), Vector Quantization (VQ) method, Polynomial Classifiers method and so on.

Preferably, the training method used in the embodiment of the present application is a genetic algorithm (GA) based on a neural network method, and the specific implementation process is as follows: By using the basic voiceprint characteristics of each participant as a group, the basic voiceprint Each voiceprint feature in the feature is used as a chromosome, and the genetic algorithm is used to perform initialization coding, individual evaluation, selection, crossover and mutation to obtain the optimal or approximate solution of the group, that is, the basic voiceprint of the participant The optimal solution of the feature. In the subsequent recognition by the voiceprint recognition model, the optimal solution closest to the target voiceprint feature is found among the optimal solutions of each basic voiceprint feature, and according to the optimal solution Correspond to the voiceprint characteristics to determine the target participants.

In this embodiment, for each participant, voiceprint feature collection is performed to obtain the basic voiceprint characteristics of the participant, and the participant identification is obtained, and based on the participant's basic voiceprint characteristics, the participant identification, and Participants 'basic voiceprint characteristics and the corresponding relationship between the participants' logos, construct the initial voiceprint recognition model, and then use the genetic algorithm to train the initial voiceprint recognition model to generate a preset voiceprint recognition model, so that When the target voiceprint feature is identified, the participant ID corresponding to the target voiceprint feature can be quickly and accurately determined, so that the participant ID of the current speaker can be automatically and accurately recorded, which is helpful to improve the efficiency and accuracy of meeting information recording.

In an embodiment, the server sends the real-time voice text to the monitoring terminal and receives the modified text fed back by the monitoring terminal. As shown in FIG. 6, after step S4 and before step S5, the conference information recording method further includes the following: step:

S71: Send the real-time voice text to the monitoring end of the conference information record.

Specifically, after performing voice recognition on the voice data and generating a real-time voice text, the real-time voice text is sent to the monitoring end, and the meeting information recording staff of the monitoring end checks the real-time voice text. If the user finds that the real-time voice text is incorrect , The monitoring terminal feeds back the modified text corresponding to the real-time voice text to the server.

For example, in a specific embodiment, the real-time voice text received by a monitoring end is "0068-Next Year's Economic and Industrial Driving Analysis". After confirmation by the user of the monitoring end, the real-time voice text is found to be incorrect. The real-time voice text was modified to "0068- Analysis of Economic and Industrial Forms for the Next Year", and the modified text was sent to the server as the modified text.

S72: If a modified text feedback is received from the monitoring end, the modified text is used to replace the real-time voice text.

Specifically, after receiving the modified text sent by the monitoring end, the server obtains the real-time voice text corresponding to the modified text and uses the modified text to replace the real-time voice text.

The correspondence between the modified text and the real-time voice text can be determined by the serial number or other identification characters in the real-time voice text, or it can be set according to actual needs, which is not specifically limited here.

Taking the modified text in step S71 as an example, it is determined that the real-time voice text corresponding to the modified text is the real-time voice text of the serial number “0068” according to the modified text “0068-Analysis of Economic and Industrial Forms in the Next Year”, and then the serial number “0068” is found. Real-time voice text "0068- Analysis of Economic and Industrial Driving in the Next Year", and replace the real-time voice text with the modified text.

In this embodiment, the real-time voice text is sent to the monitoring end of the conference information record, and the user of the monitoring end verifies the real-time voice text. If the monitoring end user verifies that there is an error in the real-time voice text, the real-time voice text is performed. Modify, generate modified text, and send the modified text to the server through the monitoring end. After receiving the modified text sent by the monitoring end, the server replaces the corresponding real-time voice text with the modified text, which improves the meeting information record. Accuracy.

It should be understood that the size of the sequence numbers of the steps in the above embodiments does not mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of this application.

In one embodiment, a conference information recording device is provided. The conference information recording device corresponds to the conference information recording method in the foregoing embodiment. As shown in FIG. 7, the conference information recording device includes a data collection module 10, a feature extraction module 20, a person determination module 30, a voice recognition module 40, and a record generation module 50. The detailed description of each function module is as follows:

The data collection module 10 is configured to obtain the voice data and the current time if a voice data transmission request of the collection channel is received;

A feature extraction module 20, configured to extract voiceprint features from voice data to obtain target voiceprint features of the voice data;

A person determination module 30 is configured to perform voiceprint recognition on a target voiceprint feature according to a preset voiceprint recognition model, and determine a target participant identification corresponding to the target voiceprint feature, wherein the voiceprint recognition model includes the basis of the participants. Voiceprint characteristics, participant identification, and the corresponding relationship between basic voiceprint characteristics and participant identification;

A voice recognition module 40, configured to perform voice recognition on voice data to generate real-time voice text;

The record generating module 50 is configured to correlate the target participant identification, the current time, and the real-time voice text to obtain the speech information of the target participant corresponding to the target participant identification.

Further, the feature extraction module 20 includes:

The voiceprint analysis unit 21 is configured to perform voiceprint analysis on the voice data to obtain an initial voiceprint;

An emphasis processing unit 22, configured to perform pre-emphasis processing on the initial voiceprint to generate an emphasis processing voiceprint with a flat spectrum;

Framing and windowing unit 23 is configured to use framing and windowing to perform framing and windowing on the weighted voiceprint to obtain a first basic voice frame;

A silent sound separating unit 24, configured to perform silent sound separation on a first basic speech frame to obtain a first target speech frame;

A feature extraction unit 25 is configured to extract voiceprint features for each first target voice frame to obtain target voiceprint features.

Further, the speech recognition module 40 includes:

A speech parsing unit 41, configured to perform speech parsing on speech data to obtain a frame set including a second basic speech frame;

The silence detection unit 42 is configured to perform silence detection on the second basic voice frame to obtain K consecutive silence frames in the second basic voice frame, where K is a natural number;

A voice segmentation unit 43 configured to divide the second basic voice frame included in the frame set into K + 1 second target voice frames according to the K silence frames;

The text conversion unit 44 is configured to convert each second target speech frame into a real-time speech text.

Further, the conference information recording device further includes:

The voiceprint collection module 61 is configured to collect voiceprint characteristics for each participant to obtain basic voiceprint characteristics for each participant;

A model building module 62 is configured to obtain the identification of each participant and construct an initial voiceprint based on the basic voiceprint characteristics of the participants, the identification of the participants, and the correspondence between the basic voiceprint characteristics of the participants and the participants' identification Recognition model

The model training module 63 is configured to train an initial voiceprint recognition model by using a genetic algorithm to generate a preset voiceprint recognition model.

Further, the conference information recording device further includes:

A text transmission module 71, configured to send real-time voice text to the monitoring end of the conference information record;

The text replacing module 72 is configured to replace the real-time voice text with the modified text if the modified text is received from the monitoring end.

For specific limitations on the conference information recording device, reference may be made to the foregoing limitation on the conference information recording method, which is not repeated here. Each module in the conference information recording device can be implemented in whole or in part by software, hardware, and a combination thereof. The above-mentioned modules may be embedded in the hardware in or independent of the processor in the computer device, or may be stored in the memory of the computer device in the form of software, so that the processor can call and execute the operations corresponding to the above modules.

In one embodiment, a computer device is provided. The computer device may be a server, and its internal structure diagram may be as shown in FIG. 8. The computer device includes a processor, a memory, a network interface, and a database connected through a system bus. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, computer-readable instructions, and a database. The internal memory provides an environment for the operation of the operating system and computer-readable instructions in a non-volatile storage medium. The database of the computer equipment is used to store the basic voiceprint characteristics and conference information records of the participants in the conference information recording method. The network interface of the computer device is used to communicate with an external terminal through a network connection. The computer-readable instructions are executed by a processor to implement a conference information recording method.

In one embodiment, a computer device is provided, including a memory, a processor, and computer-readable instructions stored on the memory and executable on the processor. The processor implements the computer-readable instructions to implement the conference information of the foregoing embodiment. The steps of the recording method are, for example, steps S1 to S5 shown in FIG. 2. Alternatively, when the processor executes the computer-readable instructions, the functions of the modules / units of the conference information recording device of the foregoing embodiment are implemented, for example, modules 10 to 50 shown in FIG. 7. To avoid repetition, we will not repeat them here.

In one embodiment, one or more non-volatile readable storage media are provided, and computer-readable instructions are stored thereon. When the computer-readable instructions are executed by one or more processors, the meeting information record of the foregoing embodiment is implemented The steps of the method, or the functions of the modules / units of the conference information recording device of the foregoing embodiment are implemented when computer-readable instructions are executed by one or more processors. To avoid repetition, details are not repeated here.

A person of ordinary skill in the art can understand that all or part of the processes in the methods of the foregoing embodiments can be implemented by using computer-readable instructions to instruct related hardware. The computer-readable instructions can be stored in one or more non-easy instructions. In the volatile readable storage medium, the computer-readable instructions, when executed, may include the processes of the embodiments of the methods described above. Wherein, any reference to the storage, storage, database, or other media used in the embodiments provided in this application may include non-volatile and / or volatile storage. Non-volatile memory may include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), dual data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous chain (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM).

Those skilled in the art can clearly understand that, for the convenience and brevity of the description, only the above-mentioned division of functional units and modules is used as an example. In practical applications, the above functions can be assigned by different functional units, Module completion, that is, dividing the internal structure of the device into different functional units or modules to complete all or part of the functions described above.

The above-mentioned embodiments are only used to describe the technical solution of the present application, but not limited thereto. Although the present application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that they can still implement the foregoing implementations. The technical solutions described in the examples are modified, or some of the technical features are equivalently replaced; and these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the application, and should be included in Within the scope of this application.

Claims (20)

A conference information recording method, characterized in that the conference information recording method includes: If a voice data transmission request for a collection channel is received, obtaining the voice data and the current time; Performing voiceprint feature extraction on the voice data to obtain a target voiceprint feature of the voice data; According to a preset voiceprint recognition model, voiceprint recognition is performed on the target voiceprint feature to determine a target participant identification corresponding to the target voiceprint feature, wherein the voiceprint recognition model includes a basic voiceprint of the participant Characteristics, participant identification, and the corresponding relationship between the basic voiceprint characteristics and the participant identification; Performing speech recognition on the speech data to generate real-time speech text; The target participant identification, the current time, and the real-time voice text are associated and recorded to obtain the speech information of the target participant corresponding to the target participant identification. The conference information recording method according to claim 1, wherein the voiceprint feature extraction of the voice data to obtain a target voiceprint feature of the voice data comprises: Performing voiceprint analysis on the voice data to obtain an initial voiceprint; Performing pre-emphasis processing on the initial voiceprint to generate an emphasis processing voiceprint with a flat frequency spectrum; Framing and windowing the frame and windowing the emphasized voiceprint to obtain a first basic speech frame; Performing silent tone separation on the first basic voice frame to obtain a first target voice frame; Voiceprint feature extraction is performed on each of the first target speech frames to obtain the target voiceprint feature. The conference information recording method according to claim 1, wherein the performing voice recognition on the voice data to generate real-time voice text comprises: Performing speech analysis on the speech data to obtain a frame set including a second basic speech frame; Performing silence detection on the second basic voice frame to obtain K consecutive silence frames in the second basic voice frame, where K is a natural number; Dividing the second basic voice frame included in the frame set into K + 1 second target voice frames according to the K silence frames; Converting each of the second target speech frames into the real-time speech text. The conference information recording method according to claim 1, characterized in that, in the voiceprint recognition of the target voiceprint feature according to a preset voiceprint recognition model, a target corresponding to the target voiceprint feature is determined Before the participants are identified, the conference information recording method further includes: Collect voiceprint features for each participant to obtain the basic voiceprint characteristics for each participant; Obtaining the identification of each participant and constructing an initial voiceprint recognition model based on the basic voiceprint characteristics of the participant, the corresponding relationship between the participant's identification, and the basic voiceprint characteristics of the participant and the participant's identification Genetic algorithm is used to train the initial voiceprint recognition model to generate the preset voiceprint recognition model. The conference information recording method according to claim 1, wherein after real-time recognition of the voice data and generation of real-time voice text, the target participant identification, the current time, and the The real-time voice and text are correlated to obtain the target participant's speech information in the current meeting, and after the steps of performing voice recognition on the voice data and generating the real-time voice text, the meeting information recording method Also includes: Sending the real-time voice text to the monitoring end of the conference information record; If a modified text fed back by the monitoring terminal is received, the real-time voice text is replaced with the modified text. A conference information recording device, characterized in that the conference information recording device includes: A data collection module, configured to obtain the voice data and the current time if a voice data transmission request of a collection channel is received; A feature extraction module, configured to perform voiceprint feature extraction on the voice data to obtain a target voiceprint feature of the voice data; A person determination module is configured to perform voiceprint recognition on the target voiceprint feature according to a preset voiceprint recognition model, and determine a target participant identification corresponding to the target voiceprint feature, where the voiceprint recognition model includes The basic voiceprint characteristics of the participants, the participant identification, and the correspondence between the basic voiceprint characteristics and the participant identification; A voice recognition module, configured to perform voice recognition on the voice data to generate real-time voice text; The record generating module is configured to record the target participant identification, the current time, and the real-time voice text in association with each other, and obtain the speech information of the target participant corresponding to the target participant identification. The conference information recording device according to claim 6, wherein the feature extraction module comprises: A voiceprint analysis unit, configured to perform voiceprint analysis on the voice data to obtain an initial voiceprint; An emphasis processing unit, configured to perform pre-emphasis processing on the initial voiceprint to generate an emphasis processing voiceprint with a flat frequency spectrum; A frame framing and windowing unit, configured to use framing and windowing to perform framing and windowing on the emphasis processing voiceprint to obtain a first basic voice frame; A silent sound separating unit, configured to perform silent sound separation on the first basic speech frame to obtain a first target speech frame; A feature extraction unit is configured to extract a voiceprint feature for each of the first target voice frames to obtain the target voiceprint feature. The conference information recording device according to claim 6, wherein the voice recognition module comprises: A speech parsing unit, configured to perform speech parsing on the speech data to obtain a frame set including a second basic speech frame; A silence detection unit, configured to perform silence detection on the second basic voice frame to obtain K consecutive silence frames in the second basic voice frame, where K is a natural number; A voice segmentation unit, configured to divide the second basic voice frame included in the frame set into K + 1 second target voice frames according to the K silence frames; A text conversion unit, configured to convert each of the second target speech frames into the real-time speech text. The conference information recording device according to claim 6, wherein the conference information recording device further comprises: The voiceprint collection module is used to collect voiceprint features for each participant to obtain the basic voiceprint characteristics of each participant; A model building module is used to obtain the identification of each participant and construct a correspondence relationship between the participant's basic voiceprint characteristics and the participants' basic voiceprint characteristics and the participant's identification based on the basic voiceprint characteristics of the participant. Initial voiceprint recognition model; A model training module is configured to train the initial voiceprint recognition model by using a genetic algorithm to generate the preset voiceprint recognition model. The conference information recording device according to claim 6, wherein the conference information recording device further comprises: A text transmission module, configured to send the real-time voice text to a monitoring end of a conference information record; A text replacement module is configured to replace the real-time voice text with the modified text if the modified text fed back by the monitoring terminal is received. A computer device includes a memory, a processor, and computer-readable instructions stored in the memory and executable on the processor, and is characterized in that the processor implements the computer-readable instructions as follows step: If a voice data transmission request for a collection channel is received, obtaining the voice data and the current time; Performing voiceprint feature extraction on the voice data to obtain a target voiceprint feature of the voice data; According to a preset voiceprint recognition model, voiceprint recognition is performed on the target voiceprint feature to determine a target participant identification corresponding to the target voiceprint feature, wherein the voiceprint recognition model includes a basic voiceprint of the participant Characteristics, participant identification, and the corresponding relationship between the basic voiceprint characteristics and the participant identification; Performing speech recognition on the speech data to generate real-time speech text; The target participant identification, the current time, and the real-time voice text are associated and recorded to obtain the speech information of the target participant corresponding to the target participant identification. The terminal device according to claim 11, wherein the voiceprint feature extraction of the voice data to obtain the target voiceprint feature of the voice data comprises: Performing voiceprint analysis on the voice data to obtain an initial voiceprint; Performing pre-emphasis processing on the initial voiceprint to generate an emphasis processing voiceprint with a flat frequency spectrum; Framing and windowing the frame and windowing the emphasized voiceprint to obtain a first basic speech frame; Performing silent tone separation on the first basic voice frame to obtain a first target voice frame; Voiceprint feature extraction is performed on each of the first target speech frames to obtain the target voiceprint feature. The terminal device according to claim 11, wherein the performing voice recognition on the voice data to generate real-time voice text comprises: Performing speech analysis on the speech data to obtain a frame set including a second basic speech frame; Performing silence detection on the second basic voice frame to obtain K consecutive silence frames in the second basic voice frame, where K is a natural number; Dividing the second basic voice frame included in the frame set into K + 1 second target voice frames according to the K silence frames; Converting each of the second target speech frames into the real-time speech text. The terminal device according to claim 11, characterized in that, in the voiceprint recognition of the target voiceprint feature according to a preset voiceprint recognition model, a target participant corresponding to the target voiceprint feature is determined Before identification, when the processor executes the computer-readable instructions, the method further includes implementing the following steps: Collect voiceprint features for each participant to obtain the basic voiceprint characteristics for each participant; Obtaining the identification of each participant and constructing an initial voiceprint recognition model based on the basic voiceprint characteristics of the participant, the corresponding relationship between the participant's identification, and the basic voiceprint characteristics of the participant and the identification of the participant; Genetic algorithm is used to train the initial voiceprint recognition model to generate the preset voiceprint recognition model. The terminal device according to claim 11, characterized in that after real-time recognition of the voice data and generation of real-time voice text, after the identification of the target participant, the current time and the real-time voice The text is associated to obtain the speech information of the target participant before the step of the conference, and after the steps of performing voice recognition on the voice data and generating real-time voice text, the processor executes the computer The readable instructions also include the following steps: Sending the real-time voice text to the monitoring end of the conference information record; If a modified text fed back by the monitoring terminal is received, the real-time voice text is replaced with the modified text. One or more non-volatile readable storage media storing computer readable instructions, characterized in that when the computer readable instructions are executed by one or more processors, the one or more processors are caused to execute The following steps: if a voice data transmission request for a collection channel is received, obtaining the voice data and the current time; Performing voiceprint feature extraction on the voice data to obtain a target voiceprint feature of the voice data; According to a preset voiceprint recognition model, voiceprint recognition is performed on the target voiceprint feature to determine a target participant identification corresponding to the target voiceprint feature, wherein the voiceprint recognition model includes a basic voiceprint of the participant Characteristics, participant identification, and the corresponding relationship between the basic voiceprint characteristics and the participant identification; Performing speech recognition on the speech data to generate real-time speech text; The target participant identification, the current time, and the real-time voice text are associated and recorded to obtain the speech information of the target participant corresponding to the target participant identification. The non-volatile readable storage medium according to claim 16, wherein the voiceprint feature extraction of the voice data to obtain the target voiceprint feature of the voice data comprises: Performing voiceprint analysis on the voice data to obtain an initial voiceprint; Performing pre-emphasis processing on the initial voiceprint to generate an emphasis processing voiceprint with a flat frequency spectrum; Framing and windowing the frame and windowing the emphasized voiceprint to obtain a first basic speech frame; Performing silent tone separation on the first basic voice frame to obtain a first target voice frame; Voiceprint feature extraction is performed on each of the first target speech frames to obtain the target voiceprint feature. The non-volatile readable storage medium according to claim 16, wherein the performing voice recognition on the voice data to generate real-time voice text comprises: Performing speech analysis on the speech data to obtain a frame set including a second basic speech frame; Performing silence detection on the second basic voice frame to obtain K consecutive silence frames in the second basic voice frame, where K is a natural number; Dividing the second basic voice frame included in the frame set into K + 1 second target voice frames according to the K silence frames; Converting each of the second target speech frames into the real-time speech text. The non-volatile readable storage medium according to claim 16, wherein the target voiceprint feature is identified by voiceprint recognition according to a preset voiceprint recognition model, and the target voiceprint is determined. Before the target participant corresponding to the feature is identified, when the computer-readable instructions are executed by one or more processors, the one or more processors further perform the following steps: Collect voiceprint features for each participant to obtain the basic voiceprint characteristics for each participant; Obtaining the identification of each participant and constructing an initial voiceprint recognition model based on the basic voiceprint characteristics of the participant, the corresponding relationship between the participant's identification, and the basic voiceprint characteristics of the participant and the identification of the participant; Genetic algorithm is used to train the initial voiceprint recognition model to generate the preset voiceprint recognition model. The non-volatile readable storage medium according to claim 16, wherein after real-time recognition of the voice data and generation of real-time voice text, the target participant identification, the current Correlate time with the real-time voice text to obtain the speech information of the target participant in the current meeting, and after the step of performing voice recognition on the voice data to generate real-time voice text, When the computer-readable instructions are executed by one or more processors, the one or more processors further perform the following steps: Sending the real-time voice text to the monitoring end of the conference information record; If a modified text fed back by the monitoring terminal is received, the real-time voice text is replaced with the modified text.

Images