WO2012163013A1 - Music query method and apparatus - Google Patents

Abstract

Provided are a music query method and apparatus. The music query method comprises: extracting a music segment to be queried from a music file to be queried, and framing the music segment to be queried; extracting the fingerprint characteristics of a frame segment contained in the music segment to be queried in order to obtain the fingerprint characteristics of the music segment to be queried; according to the fingerprint characteristics of the frame segment contained in the music segment to be queried, querying, in the fingerprint characteristics stored in a fingerprint database, the fingerprint characteristics matched with the fingerprint characteristics of the music segment to be queried, and returning a query result according to the degree of similarity between the fingerprint characteristics of the music segment to be queried and the queried fingerprint characteristics. The present invention has no requirements for the length and the starting point of the music segment to be queried, thereby improving the music query efficiency.

Description

 Music inquiry method and device

The present invention relates to the field of communications technologies, and in particular, to a music query method and apparatus.

Background technique

 With the rapid development of the Internet and digital multimedia, how to quickly and effectively perform music retrieval has become one of the research hotspots at present. The traditional text-based retrieval method can only retrieve files with annotation information, and content-based retrieval. It does not rely on the annotation information, but retrieves a piece or part of the music, which is a challenging research topic. With the development and implementation of content-based music retrieval, it will eventually bring great convenience to the retrieval of music listeners.

 A music fingerprint, referred to as an Audio Fingerprint, is defined as a sequence of audio clip features that are processed to characterize the "identity" of the music. The method of music pattern recognition and retrieval research is significantly different from the traditional music retrieval based on metadata such as song title and singer. The music does not contain all the information of a piece of music, but it can be used to identify a unique piece of music, that is, through the music pattern, the desired music can be queried from the massive data.

 The existing music query technology generally has specific requirements on the length and starting point of the song query segment, and the query efficiency is low. . Summary of the invention

 Embodiments of the present invention provide a music query method and apparatus to improve the query efficiency of music.

 An embodiment of the present invention provides a music query method, including:

 Intercepting the music piece to be queried from the music file to be queried, and framing the music piece to be queried;

Extracting fingerprint features of the segmented segments included in the music segment to be queried to obtain the Query the fingerprint characteristics of the music piece;

 Determining, according to the fingerprint feature of the segmented segment included in the music segment to be queried, a fingerprint feature matching the fingerprint feature of the music segment to be queried in the fingerprint feature stored in the fingerprint database, and according to the fingerprint of the music segment to be queried The degree of similarity between the feature and the queried fingerprint feature returns the query result.

 The embodiment of the invention further provides a music query device, including:

 An intercepting module, configured to intercept a music piece to be queried from a music file to be queried;

 a framing module, configured to framing the music segment to be queried;

 An extracting module, configured to extract a fingerprint feature of the segmented segment included in the music segment to be queried, to obtain a fingerprint feature of the music segment to be queried;

 a querying module, configured to query, according to the fingerprint feature of the segmented segment included in the music segment to be queried extracted by the extraction module, a fingerprint feature matching the fingerprint feature of the music segment to be queried in the fingerprint feature stored in the fingerprint database ;

 And a returning module, configured to return a query result according to a similarity degree between the fingerprint feature of the music piece to be queried and the fingerprint feature queried by the query module.

 In the embodiment of the present invention, the music segment to be queried is first intercepted from the music file to be queried, and the music segment to be queried is segmented, and then the fingerprint feature of the framing segment included in the music segment to be queried is extracted to obtain a fingerprint of the music segment to be queried. Finally, according to the fingerprint feature of the segmented segment included in the music segment to be queried, querying the fingerprint feature matched with the fingerprint feature of the music segment to be queried in the fingerprint feature stored in the fingerprint database, and according to the fingerprint of the music segment to be queried The degree of similarity between the feature and the queried fingerprint feature returns the query result. The embodiment of the present invention does not require the length and the starting point of the music query segment, and can improve the query efficiency of the music. DRAWINGS

The drawings used in the embodiments or the description of the prior art are briefly described. It is obvious that the drawings in the following description are some embodiments of the present invention, and are not creative to those skilled in the art. Other drawings can also be obtained from these drawings on the premise of labor.

1 is a flow chart of an embodiment of a music query method according to the present invention; 2 is a flow chart of an embodiment of a fingerprint feature extraction process of the present invention;

 3 is a schematic diagram of an embodiment of extracting a spectrum envelope and a dimension reduction process according to the present invention; FIG. 4 is a schematic structural diagram of an embodiment of a music query apparatus according to the present invention;

 FIG. 5 is a schematic structural diagram of another embodiment of a music query apparatus according to the present invention; FIG.

 FIG. 6 is a schematic structural diagram of an embodiment of a computer device according to the present invention. The technical solutions in the embodiments of the present invention are clearly and completely described in the following with reference to the accompanying drawings in the embodiments of the present invention. The embodiments are a part of the embodiments of the invention, and not all of the embodiments. All other embodiments obtained by a person of ordinary skill in the art based on the embodiments of the present invention without creative efforts are within the scope of the present invention.

 1 is a flowchart of an embodiment of a music query method according to the present invention. As shown in FIG. 1, the music query method may include:

 Step 101: Intercept the music piece to be queried from the music file to be queried, and categorize the music piece to be queried.

 Step 102: Extract fingerprint features of the segmented segments included in the music segment to be queried to obtain fingerprint features of the music segment to be queried.

 Step 103: Query, according to the fingerprint feature of the segmented segment included in the music segment to be queried, a fingerprint feature matching the fingerprint feature of the music segment to be queried in the fingerprint feature stored in the fingerprint database, and according to the fingerprint of the music segment to be queried The degree of similarity between the feature and the queried fingerprint feature returns the query result.

 Further, according to the fingerprint feature of the segmented segment included in the music segment to be queried, before the fingerprint feature matched with the fingerprint feature of the music segment to be queried is queried in the fingerprint feature stored in the fingerprint database, the known music file needs to be performed first. Framing; then extracting the fingerprint features of the segmented segments included in the known music file to obtain the fingerprint features of the known music files, and finally storing the fingerprint features of the known music files into the fingerprint database.

In this embodiment, extracting the fingerprint feature of the segmented segment included in the music segment to be queried may be: performing time-frequency conversion on the segmented segment included in the music segment to be queried, and performing frequency domain data obtained by time-frequency conversion. ; choose according to the auditory characteristics of the human ear in the frequency domain data after modeling Frequency domain data on a predetermined frequency band; extracting a spectrum envelope of frequency domain data on the predetermined frequency band; performing dimensionality reduction processing on the feature matrix obtained after extracting the spectrum envelope, and obtaining fingerprint features of the frame segment included in the music segment to be queried .

 In this embodiment, extracting the fingerprint feature of the segmented segment included in the known music file may be: performing time-frequency conversion on the segmented segment included in the known music file, and performing frequency-modulation on the frequency-domain data obtained by time-frequency conversion; Selecting frequency domain data on a predetermined frequency band in the frequency domain data after modeling according to the human auditory characteristic; extracting a spectral envelope of the frequency domain data on the predetermined frequency band; and performing dimensionality reduction on the feature matrix obtained after extracting the spectral envelope Processing, obtaining fingerprint features of the segmented segments included in the known music file.

 In this embodiment, step 103 may be: first, according to the fingerprint feature of the segmented segment included in the music segment to be queried, querying the fingerprint feature of the segmented segment included in the to-be-queried music segment in the fingerprint feature stored in the fingerprint database Matching the fingerprint feature; secondly, reading a predetermined number of fingerprint features in the fingerprint database from the location according to the location of the matched fingerprint feature stored in the fingerprint database in the belonging song, the predetermined number and the music segment to be queried include The number of the fragmented segments is the same; finally, the degree of similarity between the predetermined number of fingerprint features and the fingerprint features of all the segmented segments included in the music segment to be queried is compared, and the query result is returned according to the degree of similarity.

 The above embodiment does not require the length and starting point of the music query segment, and can improve the query efficiency of the music; and in the noisy environment, the music fingerprint query can be effectively completed, and the matching result under the song noise can be returned.

 The music query method provided by the embodiment of the present invention is described in detail in the following three aspects: the music fingerprint extraction process, the fingerprint database establishment process, and the music fingerprint query process.

 1. Music fingerprint extraction process.

 In this embodiment, the music fingerprint extraction process may include: decoding, downsampling, and fingerprint feature extraction. The following are described separately:

(1) Decoding process: Since the music file is generally encoded and compressed, before extracting the fingerprint feature of the music file, the music file is first decoded, and the music file is decoded into a waveform (Wave; hereinafter referred to as WAV) The file, the decoded music file has the same sampling rate as the original music. The sampling rate of common music files is generally 44KHz or 22KHz. (2) Downsampling process: Since the sampling rate of music files is generally high, most of the high frequency information is included, which makes it difficult to identify music files. Therefore, for the decoded music files, downsampling processing is also required. The music file to be decoded is reduced from a higher sampling rate such as 44KHz or 22KHz to a lower sampling rate. In this embodiment, the decoded music file is uniformly reduced to a sampling rate of 5KHz, and the down sampling processing is performed. The subsequent music file is converted into a file of Pulse Code Modulation (hereinafter referred to as PCM) format.

 (3) Fingerprint feature extraction process: After the decoding process and the downsampling process, the fingerprint feature extraction process is performed. As shown in FIG. 2, FIG. 2 is a flowchart of an embodiment of the fingerprint feature extraction process of the present invention, including:

 Step 201: Perform framing on the music file after the downsampling process.

 In this embodiment, when framing is performed, it is necessary to ensure partial overlap between adjacent two frames to maintain short-term stability of the signal. Specifically, the process of framing is implemented by windowing. For example: Hanning window can be used for framing, and the window size of Hanning window is 2048 points.

 Step 202: Perform time-frequency conversion on the framed segments obtained by the frame.

 Specifically, the above-mentioned frame segment is time-frequency-converted in a plurality of manners, and the implementation manner of the time-frequency conversion is not limited in this embodiment. In this embodiment, a Fast Fourier Transform (hereinafter referred to as FFT) of 2048 points is performed on the above-mentioned frame segment as an example. In this embodiment, the data repetition rate of two adjacent frames is 31/. 32, that is, the next frame is FFTed with approximately 60 new PCM data relative to the previous frame. The value obtained in step 202 is a complex number.

 Step 203: Perform modulo calculation on frequency domain data obtained by time-frequency conversion.

 Step 204: Select frequency domain data on a predetermined frequency band in the frequency domain data after the modulo according to the human ear hearing characteristic.

 In this embodiment, 33 sub-bands are selected from the frequency domain data after the modulo according to the auditory characteristics of the human ear, and the frequency distribution of the sub-bands ranges from 0 to 2.5 kHz, and the bandwidth of the above 33 sub-bands is The logarithmic domain is distributed in a linear relationship.

 Step 205: Extract a spectrum envelope of frequency domain data on the predetermined frequency band.

Specifically, the method for extracting the spectrum envelope is not limited. In this embodiment, the frequency envelope of the frequency domain data is extracted by using wavelet transform as an example. . Wavelet transform is a local transformation of space and frequency. It can perform multi-scale refinement analysis of functions or signals through operations such as scaling and translation, so that information can be extracted from signals effectively. In this embodiment, the standard Haar wavelet is used to analyze the above-mentioned frequency domain data, and only the maximum 300 wavelet coefficients (according to the absolute value of the spectral energy) are retained, and other coefficients not in the maximum 300 wavelet coefficients are quantized as "00". For each of the largest 300 wavelet coefficients, if it is a positive number, the quantization is "10", otherwise the quantization is "01".

 Step 206: Perform dimension reduction processing on the feature matrix obtained after extracting the spectrum envelope, and obtain fingerprint features of the segmented segments included in the music file.

 Since the wavelet transform is a high-dimensional 0-1 feature matrix, dimensionality reduction is required. In this embodiment, the minimum hash (MinHash) algorithm is used for the dimensionality reduction processing, that is, the arbitrary position of each 0-1 feature matrix is arbitrarily exchanged P times, and the position of the first one is recorded each time; generally speaking, at the 255th position After that, the probability of the first occurrence of 1 is very small, so 255 is uniformly taken as 255; thus, the high-dimensional 0-1 characteristic matrix is compressed into the P-dimensional eigenvalue, and each group of P-dimensional 0~255 integers is called music. A sub-grain of the pattern. In the actual implementation, P = 100 can be taken, so that the dimension of the 100-dimensional 0~255 can be obtained after the dimension reduction processing. In this embodiment, each group of 100-dimensional integers of 0 to 255 is called a sub-grain of the music pattern. .

 3 is a schematic diagram of an embodiment of extracting a spectral envelope and a dimensionality reduction process according to the present invention. As shown in FIG. 3, in this embodiment, PCM data is read in chronological order, and each frame of data is read in 60 more than the previous frame. PCM data, this process continues until it reaches the end of the PCM data. Then, the spectrum envelope of each frame of PCM data is extracted according to the method provided in step 205, and then the feature matrix obtained by extracting the spectrum envelope is subjected to dimensionality reduction processing according to the method provided in step 206, to obtain the fragmented segment included in the music file. Fingerprint feature. The fingerprint feature of the segmented segment is referred to as a sub-grain, and the fingerprint feature of the music file is referred to as a tune. As can be seen from FIG. 3, the tune is a sequence of sub-tones, and the sequence of sub-tones in the sequence The sequence reflects the sequential relationship of the sub-frame segments corresponding to the sub-tones in time.

 2. The fingerprint database establishment process.

 (1) Extract the fingerprint features of the known music files that need to be stored in accordance with the method provided in the above musical fingerprint extraction process.

(2) Two data tables are stored in the fingerprint database, one is an index table, and the other is a music sheet. Each entry in the index table stores a child note and the child note in the fingerprint database An identification, and a specific time position of the sub-grain in the song to which it belongs. Each entry in the music sheet stores the music pattern of a song, that is, all the child music patterns contained in the song.

 (3) Music storage: The specific operation is to record each sub-grain of each song into the index table in turn, and assign a unique identifier in the fingerprint database for each sub-grain in the index table, and each song contains All child music patterns are entered into the music sheet.

 3. Music fingerprint inquiry process.

 In the embodiment of the present invention, in the process of the music fingerprint query, the fingerprint feature of the music segment to be queried is first extracted according to the method provided in the above music fingerprint extraction process, and then the fingerprint feature of the segmented segment included in the music segment to be queried is first The fingerprint feature matching the fingerprint feature of the segmented segment included in the music segment to be queried is searched in the index table, and then the fingerprint matching the above is found according to the identifier of the matched fingerprint feature saved in the index table. The entry corresponding to the feature, according to the position of the matching fingerprint feature saved in the index table in the belonging song, reads a predetermined number of sub-categories from the position corresponding to the fingerprint pattern and the matching fingerprint feature from the position A note pattern, wherein the predetermined number is the same as the number of frame segments included in the music piece to be queried. Finally, the degree of similarity between the read predetermined number of sub-letter patterns and the fingerprint features of all the sub-frame segments included in the music segment to be queried is compared, and the query result list is returned according to the degree of similarity.

 In summary, the music query method provided by the embodiment of the present invention has the following advantages:

 1. The compression ratio of the music pattern can reach more than 100 times of compression, the compression ratio is large, and the representation is strong. In the noisy environment, the wavelet transform can be used to extract the characteristics of the noise details, and the high-energy information part of the spectrogram is hash-compressed, so that one frame of data is compressed from the original 8192 points to 100 bytes; The data is reduced to a few hundredths of the original data; therefore, the feature compression ratio is large and representative.

 2, music pattern design has a certain degree of noise resistance.

On the basis of extracting the energy spectrum, wavelet transform is further performed, and only the largest 300 wavelet coefficients with relatively obvious amplitude are extracted, thereby avoiding certain noise signal effects; the feature matrix is processed by the minimum hash algorithm, which is up to 8192. The feature of the dimension is reduced in dimension, and the feature can be similarly calculated by a simple comparison. When the database linked list is manufactured, the local sensitive hash is introduced in consideration of the variation characteristics of the local features of the music, and the applicability is strong. , greatly reducing the range of candidate music patterns. Since the noise resistance tolerance is taken into account in the music pattern extraction stage, it does not contain a special denoising system, so the final noise-free music pattern has a certain Noise resistance.

 3. Perform different queries for different query requirements.

 According to the practical application considerations, the embodiment of the present invention can return the query result list according to the degree of similarity.

 4. Embodiments of the present invention may also perform similarity comparison and metrics of overlapping portions on similar pieces of music.

 Since the music patterns extracted by the embodiment of the present invention are sequential in time, it is convenient to know the source of the two segments and the position in the belonging song, thereby judging the similarity of the two similar music segments and The proportion of overlap. The sequence of the fingerprint data in the fingerprint data and the efficiency of the query ensure the realization of such requirements.

 A person skilled in the art can understand that all or part of the steps of implementing the above method embodiments may be completed by using hardware related to program instructions, and the foregoing program may be stored in a computer readable storage medium, and the program is executed when executed. The foregoing steps include the steps of the foregoing method embodiments; and the foregoing storage medium includes: a medium that can store program codes, such as a ROM, a RAM, a magnetic disk, or an optical disk.

 4 is a schematic structural diagram of an embodiment of a music query apparatus according to the present invention. As shown in FIG. 4, the music query apparatus may include: an intercepting module 41, a framing module 42, an extracting module 43, a querying module 44, and a returning module 45.

 The intercepting module 41 is configured to intercept the music segment to be queried from the music file to be queried; the framing module 42 is configured to framing the music segment to be queried;

 The extracting module 43 is configured to extract a fingerprint feature of the segmented segment included in the music segment to be queried to obtain a fingerprint feature of the music segment to be queried;

 The querying module 44 is configured to query, according to the fingerprint feature of the frame segment included in the music segment to be queried extracted by the extraction module 43 , a fingerprint feature matching the fingerprint feature of the music segment to be queried in the fingerprint feature stored in the fingerprint database;

 The returning module 45 is configured to return the query result according to the degree of similarity between the fingerprint feature of the music piece to be queried and the fingerprint feature queried by the query module 44.

The music inquiry device does not require the length and the starting point of the music query segment, and can improve the query efficiency of the music; and in the noisy environment, the music fingerprint query can be effectively completed, and the matching result under the song noise can be returned. Figure 5 is a schematic diagram of another embodiment of the music query device of the present invention, as shown in Figure 5, the music query device may further include: a storage module 46;

 In this embodiment, the framing module 42 is further configured to framing the known music file; the extracting module 43 is further configured to extract the fingerprint feature of the framing segment included in the known music file to obtain the known music. The fingerprint characteristics of the document;

 The storage module 46 is configured to store the fingerprint feature of the above-mentioned known music file obtained by the extraction module 43 into the fingerprint database.

 Specifically, the extraction module 43 may include: a conversion submodule 43 1. a module submodule 432, a selection submodule 433, an envelope extraction submodule 434, and a dimension reduction submodule 435.

 The conversion sub-module 43 1 is configured to perform time-frequency conversion on the frame segment included in the music segment to be queried;

 The module submodule 432 is configured to perform frequency domain data obtained by time-frequency conversion, and the selection sub-module 433 is configured to select a frequency domain on a predetermined frequency band in the frequency domain data obtained by the module sub-module 432 according to the human ear hearing characteristic. Data

 An envelope extraction submodule 434, configured to extract a spectrum envelope of frequency domain data on the predetermined frequency band;

 The dimension reduction sub-module 435 is configured to perform a dimensionality reduction process on the feature matrix obtained by extracting the spectrum envelope by the envelope extraction sub-module 434, and obtain the fingerprint feature of the segmented segment included in the music segment to be queried.

 Specifically, in this embodiment, the query module 44 may include: a feature query sub-module 441 and a feature reading sub-module 442;

 The feature query sub-module 441 is configured to query, according to the fingerprint feature of the segmented segment included in the music segment to be queried, a fingerprint feature stored in the fingerprint database to match a fingerprint feature of the segmented segment included in the to-be-queried music segment. Fingerprint feature

 The feature reading sub-module 442 is configured to read, according to the location of the matched fingerprint feature stored in the fingerprint database, a predetermined number of fingerprint features in the fingerprint database, the predetermined number and the music to be queried. The fragment contains the same number of fragmented segments.

In this embodiment, the returning module 45 may compare the degree of similarity between the predetermined number of fingerprint features read by the feature reading sub-module 442 and the fingerprint features of all the segmented segments included in the music segment to be queried, and return the query result according to the similarity degree. . The music inquiry device does not require the length and the starting point of the music query segment, and can improve the query efficiency of the music; and in the noisy environment, the music fingerprint query can be effectively completed, and the matching result under the song noise can be returned.

 FIG. 6 is a schematic structural diagram of an embodiment of a computer device according to the present invention. The computer device in this embodiment can implement the function of the music query device in the embodiment shown in FIG. 4 or FIG. 5, as shown in FIG. The system may include: a central processing unit (hereinafter referred to as: CPU) 61, a bus control logic 62, a system bus 63, a memory 64, an interface 65, and an input/output (I/O) subsystem 66. Wherein the I/O subsystem 66 includes an I/O device 661 and a memory 662.

 In this embodiment, the CPU 61 is configured to intercept a music segment to be queried from the music file to be queried, perform framing on the music segment to be queried, and extract a fingerprint feature of the framing segment included in the music segment to be queried to obtain the above. Querying the fingerprint feature of the music segment to be queried, according to the fingerprint feature of the segmented segment included in the extracted music segment, querying the fingerprint feature stored in the fingerprint database to match the fingerprint feature of the music segment to be queried, and according to Querying the similarity between the fingerprint feature of the music piece and the queried fingerprint feature returns the query result; the CPU 61 in this embodiment can implement the intercepting module 41, the framing module 42, and the extraction in the embodiment shown in FIG. 4 or FIG. The functions of module 43 and query module 44.

 The fingerprint database is stored in the memory 662. Specifically, the CPU 61 returns the query result: the CPU 61 sends the query result to the bus control logic 62, and the query result is passed by the bus control logic 62 through the system bus 63 and the interface 65. The I/O device 661 sends the query result to the I/O device 661. In addition, before the I/O device 661 sends the query result, the query result may be cached in the memory 64 first. That is, in this embodiment, the CPU 61, the bus control logic 62, the system bus 63, the memory 64, the interface 65, and the I/O device 661 collectively complete the return module 45 of the embodiment shown in FIG. 4 or FIG. 5 of the present invention. Features.

 Further, the CPU 61 may further framing the known music files to extract the fingerprint features of the segmented segments included in the known music files to obtain the fingerprint features of the known music files.

In this embodiment, the memory 662 is configured to save the fingerprint database, and store the fingerprint feature of the known music file obtained by the CPU 61 into the fingerprint database. The memory in this embodiment 662 can implement the functions of memory module 46 in the embodiment of FIG. 5 of the present invention.

 The above computer device does not require the length and starting point of the music query segment, and can improve the query efficiency of the music; and in the noisy environment, the music fingerprint query can be effectively completed, and the matching result under the song noise can be returned.

 A person skilled in the art can understand that the drawings are only a schematic diagram of a preferred embodiment, and the modules or processes in the drawings are not necessarily required to implement the invention.

 Those skilled in the art can understand that the modules in the apparatus in the embodiments may be distributed in the apparatus of the embodiment according to the embodiment description, or the corresponding changes may be located in one or more apparatuses different from the embodiment. The modules of the above embodiments may be combined into one module, or may be further split into a plurality of sub-modules.

 It should be noted that the above embodiments are only for explaining the technical solutions of the present invention, and are not intended to be limiting; although the present invention has been described in detail with reference to the foregoing embodiments, it will be understood by those skilled in the art that: The technical solutions described in the foregoing embodiments are modified, or some of the technical features are equivalently replaced. The modifications and substitutions do not depart from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

Claim A music query method, comprising:  Intercepting the music piece to be queried from the music file to be queried, and framing the music piece to be queried;  Extracting fingerprint features of the segmented segments included in the music segment to be queried to obtain fingerprint features of the music segment to be queried;  Determining, according to the fingerprint feature of the segmented segment included in the music segment to be queried, a fingerprint feature matching the fingerprint feature of the music segment to be queried in the fingerprint feature stored in the fingerprint database, and according to the fingerprint of the music segment to be queried The degree of similarity between the feature and the queried fingerprint feature returns the query result.  The method according to claim 1, wherein the querying the fingerprint feature of the to-be-queried music segment in the fingerprint feature stored in the fingerprint database according to the fingerprint feature of the segmented segment included in the music segment to be queried Before the fingerprint feature matching the fingerprint feature, the method further includes: framing the known music file;  Extracting fingerprint features of the segmented segments included in the known music file to obtain fingerprint features of the known music files;  The fingerprint features of the known music files are stored to the fingerprint database.  The method according to claim 1, wherein the extracting the fingerprint features of the segmented segments included in the music segment to be queried comprises:  And performing time-frequency conversion on the frame segment included in the music segment to be queried, and performing frequency modulo calculation on the frequency domain data obtained by the time-frequency conversion;  Selecting frequency domain data on a predetermined frequency band in the frequency domain data after modulo according to human auditory characteristics;  Extracting a spectral envelope of frequency domain data on the predetermined frequency band;  Performing dimensionality reduction on the feature matrix obtained after extracting the spectral envelope, obtaining fingerprint features of the segmented segments included in the to-be-queried music segment.  4. The method according to claim 2, wherein the extracting the fingerprint features of the segmented segments included in the known music file comprises: Performing time-frequency conversion on the frame segment included in the known music file, and performing frequency domain data obtained by the time-frequency conversion; Selecting frequency domain data on a predetermined frequency band in the frequency domain data after modulo according to human auditory characteristics;  Extracting a spectral envelope of frequency domain data on the predetermined frequency band;  The feature matrix obtained after extracting the spectrum envelope is subjected to dimensionality reduction processing to obtain fingerprint features of the segmented segments included in the known music file.  The method according to claim 3 or 4, wherein the querying the music to be queried in the fingerprint feature stored in the fingerprint database according to the fingerprint feature of the segmented segment included in the music segment to be queried The fingerprint feature matching the fingerprint feature of the segment, and returning the query result according to the similarity between the fingerprint feature of the music segment to be queried and the queried fingerprint feature includes:  And querying, according to the fingerprint feature of the segmented segment included in the music segment to be queried, a fingerprint feature matching the fingerprint feature of the segmented segment included in the music segment to be queried in the fingerprint feature stored in the fingerprint data library;  And reading, according to the location of the matched fingerprint feature stored in the fingerprint database, a predetermined number of fingerprint features in the fingerprint database from the location, the predetermined number and the music segment to be queried Contains the same number of fragmented segments;  Comparing the degree of similarity between the predetermined number of fingerprint features and the fingerprint features of all the sub-frame segments included in the music segment to be queried, and returning the query result according to the similarity degree.  6. A music query device, comprising:  An intercepting module, configured to intercept a music segment to be queried from a music file to be queried; a framing module, configured to framing the music segment to be queried;  An extracting module, configured to extract a fingerprint feature of the segmented segment included in the music segment to be queried, to obtain a fingerprint feature of the music segment to be queried;  a querying module, configured to query, according to the fingerprint feature of the segmented segment included in the music segment to be queried extracted by the extraction module, a fingerprint feature matching the fingerprint feature of the music segment to be queried in the fingerprint feature stored in the fingerprint database ;  And a returning module, configured to return a query result according to a similarity degree between the fingerprint feature of the music piece to be queried and the fingerprint feature queried by the query module. The device according to claim 6, further comprising: a storage module; the framing module is further configured to perform framing on the known music file; The extracting module is further configured to extract fingerprint features of the segmented segments included in the known music file to obtain fingerprint features of the known music files;  The storage module is configured to store fingerprint features of the known music files obtained by the extraction module to the fingerprint database.  8. The apparatus according to claim 6 or 7, wherein the extraction module comprises:  a conversion submodule, configured to perform time-frequency conversion on the frame segment included in the music segment to be queried;  a module for modulo frequency domain data obtained by the time-frequency conversion; a submodule for selecting a frequency in a predetermined frequency band in frequency domain data obtained by the modulo submodule according to human auditory characteristics Domain data  An envelope extraction submodule, configured to extract a spectrum envelope of frequency domain data on the predetermined frequency band; and a dimension reduction submodule, configured to perform dimensionality reduction on a feature matrix obtained by extracting a spectrum envelope by the envelope extraction submodule Processing, obtaining fingerprint features of the segmented segments included in the music segment to be queried.  The device according to claim 8, wherein the query module comprises: a feature query sub-module, configured to store in the fingerprint database according to a fingerprint feature of a segmented segment included in the music segment to be queried Querying, in the fingerprint feature, a fingerprint feature that matches a fingerprint feature of the segmented segment included in the music segment to be queried;  a feature reading submodule, configured to read a predetermined number of fingerprint features in the fingerprint database from the location according to the location of the matched fingerprint feature stored in the fingerprint database in the belonging song, the predetermined The number is the same as the number of fragmented segments included in the music segment to be queried.  10. Apparatus according to claim 9 wherein:  The returning module is specifically configured to compare the degree of similarity between a predetermined number of fingerprint features read by the feature reading sub-module and fingerprint features of all the segmented segments included in the music segment to be queried, and return according to the similarity degree search result.