CN115862653A

CN115862653A - Audio denoising method and device, computer equipment and storage medium

Info

Publication number: CN115862653A
Application number: CN202211507087.2A
Authority: CN
Inventors: 杨洁琼; 罗亚明; 王亚新; 曾德林
Original assignee: Industrial and Commercial Bank of China Ltd ICBC
Current assignee: Industrial and Commercial Bank of China Ltd ICBC
Priority date: 2022-11-29
Filing date: 2022-11-29
Publication date: 2023-03-28

Abstract

The application relates to an audio denoising method, an audio denoising device, computer equipment and a storage medium, and relates to the technical field of computers. Can be used in the field of financial technology or other related fields. The method comprises the following steps: the audio data coding sequence is shifted and transformed to obtain a shift sequence set, the shift sequences in the shift sequence set are divided to obtain a subsequence set corresponding to the shift sequence set, measurement analysis of target character strings is further performed on the subsequences in the subsequence set to obtain statistical characteristics of the subsequences, and denoising processing can be directly performed on the audio data based on the statistical characteristics of the subsequences. By adopting the method, the noise characteristic is not required to be labeled in advance, and the audio data can be subjected to denoising treatment directly.

Description

Audio denoising method and device, computer equipment and storage medium

Technical Field

The present application relates to the field of computer technologies, and in particular, to an audio denoising method and apparatus, a computer device, and a storage medium.

Background

With the development of the noise processing field, audio denoising technology has emerged, in which low-frequency noise can be removed by using, for example, a high-pass filter, or continuous sound in some frequency bands can be removed by using some notch filters.

However, in the current audio denoising method, the noise characteristics need to be labeled in advance, for example, the frequency band where the noise appears is determined, or the spectral energy of the noise is predetermined, and then the noise processing is performed based on the determined noise characteristics; in this way, the processing effect is not ideal for the use scene without marking the noise characteristics in advance.

Disclosure of Invention

In view of the foregoing, it is desirable to provide an audio denoising method, an audio denoising apparatus, a computer device, and a storage medium, which can directly perform denoising processing on audio data without labeling noise features in advance.

In a first aspect, the present application provides an audio denoising method, including:

carrying out shift transformation on the coding sequence of the audio data to obtain a shift sequence set;

dividing the shift sequences in the shift sequence set to obtain a subsequence set corresponding to the shift sequences;

measuring and analyzing the subsequences in the subsequence set to obtain the statistical characteristics of the subsequences;

and denoising the audio data according to the statistical characteristics of the subsequences.

In one embodiment, dividing the shifted sequences in the shifted sequence set to obtain a subsequence set corresponding to the shifted sequences includes:

dividing the shift sequences in the shift sequence set based on the first set length to obtain a window sequence set corresponding to the shift sequences;

based on the second set length, dividing the window sequences in the window sequence set corresponding to the shift sequences to obtain a subsequence set corresponding to the shift sequences; wherein, the first set length is greater than the second set length.

In one embodiment, the performing a measure analysis on the target string on the subsequences in the subsequence set to obtain statistical characteristics of the subsequences includes:

for each subsequence in the subsequence set, determining the statistical characteristics of the subsequence according to the length of the subsequence, the number of first target character strings in the subsequence and the number of second target character strings in the subsequence; wherein the first target string is different from the target string.

In one embodiment, determining the statistical characteristics of the subsequence according to the length of the subsequence, the number of first target strings in the subsequence, and the number of second target strings in the subsequence comprises:

taking the ratio of the length of the subsequence to the number of first target character strings in the subsequence as a first probability distribution in the statistical characteristics of the subsequence;

and taking the ratio of the length of the subsequence to the number of second target character strings in the subsequence as a second probability distribution in the statistical characteristics of the subsequence.

In one embodiment, denoising the audio data according to the statistical characteristics of the sub-sequences includes:

clustering the subsequences according to the statistical characteristics;

determining a noise distribution area in the audio data according to the number of the various subsequences;

and denoising the noise distribution area.

determining a first probability distribution of the sequence of shifts in the set of shift sequences based on the first probability distribution in the statistical features of the subsequences;

obtaining a first probability distribution of the shift sequence set according to the first probability distribution of the shift sequences;

determining a second probability distribution of the shift sequence in the shift sequence set according to the second probability distribution in the statistical characteristics of the subsequences;

obtaining a second probability distribution of the shift sequence set according to the second probability distribution of the shift sequence;

and denoising the audio data according to the first probability distribution and the second probability distribution of the displacement sequence set.

In a second aspect, the present application further provides an audio denoising apparatus, including:

the data shifting module is used for shifting and transforming the coded sequence of the audio data to obtain a shifting sequence set;

the sequence dividing module is used for dividing the shift sequences in the shift sequence set to obtain a subsequence set corresponding to the shift sequences;

the measure analysis module is used for carrying out measure analysis on the target character string on the subsequences in the subsequence set to obtain the statistical characteristics of the subsequences;

and the audio denoising module is used for denoising the audio data according to the statistical characteristics of the subsequences.

In a third aspect, the present application further provides a computer device, which includes a memory and a processor, wherein the memory stores a computer program, and the processor implements the following steps when executing the computer program:

In a fourth aspect, the present application also provides a computer readable storage medium having a computer program stored thereon, the computer program when executed by a processor implementing the steps of:

carrying out shift transformation on the coded sequence of the audio data to obtain a shift sequence set;

In a fifth aspect, the present application also provides a computer program product comprising a computer program which, when executed by a processor, performs the steps of:

According to the audio denoising method, the audio denoising device, the computer equipment and the storage medium, the shift sequence set can be obtained by performing shift transformation on the coding sequence of the audio data, the shift sequences in the shift sequence set are divided, the subsequence set corresponding to the shift sequence can be obtained, further, the measure analysis of the target character string is performed on the subsequences in the subsequence set, the statistical characteristics of the subsequences are obtained, and the audio data can be directly denoised based on the statistical characteristics of the subsequences. According to the scheme, the coding sequence of the audio data needing denoising is directly subjected to displacement transformation, division, measure analysis and the like to obtain the statistical characteristics, and the noise possibly existing in the audio data can be identified and denoised based on the statistical characteristics; in addition, according to the method and the device, the statistical characteristics introduced by the audio data displacement transformation, division, measure analysis and the like are processed, the anti-interference capacity is strong, and the audio denoising result is more accurate.

Drawings

FIG. 1 is a diagram of an embodiment of an application environment of an audio denoising method;

FIG. 2 is a flow diagram of an embodiment of a method for audio denoising;

FIG. 3 is a flow diagram illustrating division of a shifted sequence according to an embodiment;

FIG. 4 is a schematic flow chart illustrating a process for determining statistical characteristics of subsequences in one embodiment;

FIG. 5 is a schematic flow chart illustrating an embodiment of audio denoising based on statistical features;

FIG. 6 is a schematic flow chart illustrating audio denoising based on statistical features according to another embodiment;

FIG. 7 is a flowchart illustrating an audio denoising method according to another embodiment;

FIG. 8 is a block diagram showing the structure of an audio denoising apparatus according to an embodiment;

FIG. 9 is a block diagram of the structure of a sequence partitioning module in one embodiment;

FIG. 10 is a block diagram of the structure of the measure analysis module in one embodiment;

FIG. 11 is a diagram illustrating an internal structure of a computer device in one embodiment.

Detailed Description

In order to make the objects, technical solutions and advantages of the present application more apparent, the present application is described in further detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the present application and are not intended to limit the present application.

The audio denoising method provided by the embodiment of the application can be applied to the application environment shown in fig. 1. Wherein the terminal 102 communicates with the server 104 via a network. The data storage system may store data that the server 104 needs to process, such as audio data to be denoised, and the like. The data storage system may be integrated on the server 104 or may be placed on the cloud or other network server. The audio denoising method provided by the embodiment of the application can be applied to the server 104, can also be applied to the terminal 102, and can also be realized through the interaction between the terminal 102 and the server 104. Illustratively, the server 104 performs shift transformation on a coding sequence of the audio data to obtain a shift sequence set, divides shift sequences in the shift sequence set to obtain a subsequence set corresponding to the shift sequences, performs measurement analysis on target character strings on subsequences in the subsequence set to obtain statistical characteristics of the subsequences, and can perform denoising processing on the audio data according to the statistical characteristics of the subsequences; further, the server 104 may send the processed audio data to the terminal 102, so that the terminal 102 performs subsequent processing on the processed audio data based on the actual usage scenario, such as sound source identification. The terminal 102 may be, but is not limited to, various personal computers, notebook computers, smart phones, tablet computers, and the like. The server 104 may be implemented as a stand-alone server or a server cluster comprised of multiple servers.

The existing audio denoising method needs to label noise characteristics in advance, for example, determine a frequency band where noise appears, or determine noise characteristics such as spectral energy of noise in advance to perform denoising processing. For example, when the in-vehicle call recording is denoised, the noise spectrum energy is determined by analyzing other in-vehicle call recordings, and then the denoising processing is performed on the audio data to be denoised on the basis of the determined noise energy spectrum. In this way, the processing effect is not ideal for the use scene without marking the noise characteristics in advance.

Based on this, in one embodiment, as shown in fig. 2, an audio denoising method is provided, which is described by taking the method as an example applied to the server in fig. 1, and includes the following steps:

s201, carrying out shift transformation on the coded sequence of the audio data to obtain a shift sequence set.

The audio data is the audio data to be denoised; the coding sequence is a sequence obtained by coding the audio data according to a set coding rule. Alternatively, the encoding rule may be determined according to the denoising result of other audio data. For example, the encoding rule is binary encoding, and the encoding sequence obtained by encoding the audio data is a bit sequence.

Alternatively, the shift transformation refers to a manner of changing the sequence of the encoded sequence by shifting, and in the present embodiment, the shifting manner may be determined according to the encoding manner of the encoded sequence, the source and the usage scenario of the audio data, and the like. The shifting modes include shifting to the right, shifting to the left, and cross-shifting to the left and right, etc. For example, the coded sequence obtained by coding the audio data by the binary coding rule may be shifted by a set length to the right, and the shifted characters are added to the start position of the coded sequence in the original order.

One or more shifted sequences may be included in the set of shifted sequences. Optionally, a plurality of different shift lengths may be adopted to perform shift transformation on the coding sequence, respectively, to obtain a shift sequence set. In one embodiment, the different shift lengths may form an arithmetic series, for example, the first shift length is 1, the second shift length is 2, \ 8230, the nth shift length is n, etc. For example, in the case of shift transformation to the right shift mode, assuming that the coding sequence obtained by encoding the audio data to be denoised by the two-level system is 1011011010010100110, one shift length m =4, and one shift length m =8, after the shift transformation, the shifted sequence 0110101101001010 and the shifted sequence 1010011010110100 can be obtained.

Illustratively, the length of the coded sequence of the audio data is too long, and the length of the sequence is at most 2 ⁿ Therefore, when performing shift conversion, the length m =2 can be selected ⁿ Is the shift length. Wherein the shift length is less than the length of the coding sequence. Further, the length of the shifted sequence is equal to the length of the coding sequence.

S202, dividing the shift sequences in the shift sequence set to obtain a subsequence set corresponding to the shift sequences.

Optionally, the partition rule may be determined based on the coding mode, the length of the coding sequence, and the like; and dividing each shift sequence in the shift sequence set according to a division rule. Specifically, for each shift sequence in the shift sequence set, the shift sequence is divided according to a division rule to obtain a subsequence set corresponding to the shift sequence. For example, the division length in the division rule may be a number by which the length of the shift sequence is divisible, for convenience of subsequent processing.

For example, assuming that a shift sequence is 0110101101001010, the division length can be 4 according to the length of the shift sequence, and the shift sequence is divided from left to right to obtain a subsequence set of {0110, 1011, 0100, 1010}.

Further, because the length of the shifted sequence is too long, a sub-sequence set can be obtained by performing multi-level division on the shifted sequence, that is, the division times can be more than one time. In practical applications, in order to ensure the accuracy of the denoising result and the high denoising efficiency, the shift sequence may be divided by two divisions, and the specific division process will be described in detail in the following embodiments.

S203, performing measure analysis on the subsequences in the subsequence set to obtain the statistical characteristics of the subsequences.

In this embodiment, the target character string may be determined according to the code characters included in the code sequence. For example, the code characters included in the code sequence include 1, 0, 10, 11, 01, 00, and the like, and since the character string changes when the shift change is performed, the analysis can be performed on the character strings 1 and 01 in this embodiment. The metric analysis may also be referred to as statistical analysis for counting the occurrence of the target string in the sub-sequence.

Specifically, for each subsequence in the subsequence set corresponding to each shifted sequence, the number of the target character strings appearing in the subsequence can be counted, and the counted number is converted according to a set conversion rule, so as to obtain the statistical characteristics of the subsequence. For example, the statistical characteristic of the subsequence may be determined based on a ratio of the counted number to a length of the shifted sequence to which the subsequence corresponds.

It should be noted that, in this embodiment, the number of the target character strings may be one or more; to ensure the accuracy of the denoising process, in one embodiment, the number of target strings is at least two. Furthermore, for each sub-sequence, a measure analysis of each target string is performed on the sub-sequence to determine the statistical characteristics of the sub-sequence.

And S204, denoising the audio data according to the statistical characteristics of the subsequences.

Specifically, statistical analysis can be performed on the statistical characteristics of all the subsequences to obtain the noise distribution condition; and mapping the subsequence with the noise to an encoding sequence of the audio data based on the obtained noise distribution condition, accurately obtaining a frequency band where the noise is located based on the encoding sequence, and further performing denoising processing on the determined noise frequency band. The denoising process may adopt operations such as filtering process.

Illustratively, a neural network model capable of performing audio denoising processing can be obtained by training based on historical audio denoising data, and the audio data and the obtained noise distribution condition are input into the trained neural network model, so that the model can perform automatic denoising processing.

In the audio denoising method, a shift sequence set can be obtained by shifting and transforming the coding sequence of the audio data, the shift sequences in the shift sequence set are divided to obtain a subsequence set corresponding to the shift sequence set, further, the subsequences in the subsequence set are subjected to measurement analysis of the target character string to obtain the statistical characteristics of the subsequences, and the audio data can be directly denoised based on the statistical characteristics of the subsequences. According to the scheme, the coding sequence of the audio data needing denoising is directly subjected to displacement transformation, division, measure analysis and the like to obtain the statistical characteristics, and the noise possibly existing in the audio data can be identified and denoised based on the statistical characteristics; in addition, according to the method and the device, the statistical characteristics introduced by the processing of audio data such as shift transformation, division and measure analysis are adopted, the anti-interference capability is high, and the audio denoising result is accurate.

Fig. 3 is a schematic flow chart of dividing a shift sequence in an embodiment, and this embodiment further elaborates S202, that is, an operation of dividing the shift sequence into sub-sequence sets on the basis of the above embodiment, and specifically includes the following steps:

s301, based on the first set length, dividing the shift sequences in the shift sequence set to obtain a window sequence set corresponding to the shift sequences.

The first set length is a length that is set based on the encoding method of the audio data, the length of the shift sequence, and the like, and that can reasonably divide the shift sequence. Further, the first set length is smaller than the length of the shifted sequence.

Specifically, for each shift sequence in the shift sequence set, the shift sequence may be divided according to a left-to-right sequence with a first set length as a dividing unit, and the sequences obtained after the division are arranged according to a time sequence to obtain a window sequence set corresponding to the shift sequence, that is, one shift sequence corresponds to one window sequence set. Further, at least two window sequences are included in one window sequence set.

S302, based on the second set length, dividing the window sequences in the window sequence set corresponding to the shift sequences to obtain a subsequence set corresponding to the shift sequences.

The second set length is a length that is set based on the coding method of the audio data, the length of the window sequence, and the like, and can reasonably divide the window sequence, corresponding to the first set length. Wherein the second set length is smaller than the first set length.

Specifically, for each shift sequence, the window sequences in the window sequence set corresponding to the shift sequence may be divided in order from left to right by using the second set length as a division unit, so as to obtain a sub-sequence set corresponding to the shift sequence. Optionally, one window sequence may be divided into a plurality of subsequences, that is, one window sequence corresponds to one subsequence set. Further, one shifted sequence includes a plurality of window sequences, and one shifted sequence corresponds to a plurality of sub-sequence sets.

It can be understood that, in this embodiment, by dividing the shift sequence for multiple times, the obtained subsequence has a finer granularity, and further, the characteristics of the audio data can be more accurately analyzed, that is, the noise position can be more accurately located, so that the accuracy of the audio denoising processing is improved.

When the shifted sequence is divided, the division of the sub-sequence length may be uneven, and errors caused by the division may be avoided. On the basis of the above-described embodiment, the length of the subsequence is introduced in the process of determining the statistical characteristics of the subsequence. Optionally, for each subsequence in the subsequence set corresponding to each shifted sequence, the statistical characteristics of the subsequence may be determined according to the length of the subsequence, the number of first target character strings in the subsequence, and the number of second target character strings in the subsequence; wherein the first target string is different from the target string.

Optionally, as shown in fig. 4, the specific implementation process includes the following steps:

s401, taking the ratio of the length of the subsequence to the number of the first target character strings in the subsequence as a first probability distribution in the statistical characteristics of the subsequence.

The first target character string is a character string which is obtained based on an audio data coding mode and can measure the character distribution condition in the subsequence; the first probability distribution refers to a frequency of occurrence of the first target string in the subsequence.

Alternatively, in the case where the coding sequence is a bit sequence, the first target string may be set to 1, and the first probability distribution may be set to P. Further, based on the length K of the target subsequence and the number M1 of first target strings in the subsequence, a first probability distribution P = K/M1 in the statistical characteristics of the subsequence can be obtained.

S402, taking the ratio of the length of the subsequence to the number of the second target character strings in the subsequence as a second probability distribution in the statistical characteristics of the subsequence.

The second target character string refers to another character string which is obtained based on the coding mode of the audio data and can measure the character distribution condition in the subsequence; the second probability distribution refers to the frequency with which the second target string occurs in the subsequence.

Alternatively, in the case where the coding sequence is a bit sequence, the second target character string may be set to 01, and the second probability distribution may be set to Q. Further, based on the length K of the target subsequence and the number M2 of second target strings in the subsequence, a second probability distribution Q = K/M2 in the statistical characteristics of the subsequence can be obtained.

In this embodiment, the statistical characteristics are extracted by using a statistical method, that is, the probability distribution of the target character string in the subsequence is calculated to obtain the statistical characteristics of the subsequence, so that the anti-interference capability of audio denoising is improved.

Fig. 5 is a schematic flow chart of performing audio denoising based on statistical characteristics in an embodiment, and based on the above embodiment, this embodiment further explains S204 in detail, which specifically includes the following steps:

and S501, clustering the subsequences according to the statistical characteristics.

Optionally, in this embodiment, subsequences with the same statistical characteristics are used as a class. Further, in the case where the statistical features of the subsequences include a first probability distribution and a second probability distribution, the subsequences having the same first probability distribution and the same second probability distribution may be taken as a class. I.e. a class of statistical features that is essentially of one type.

And S502, determining a noise distribution area in the audio data according to the number of the subsequences of each type.

It will be appreciated that the statistical features of the noise data in the audio data are more and less than the statistical features of the normal audio data, which results in a smaller number of statistical feature classes when clustering the statistical features of the sub-sequences, where the sub-sequences with the statistical features are the sequences of the noise.

Optionally, the number of the subsequences in each category obtained in S501 may be counted, and the categories are sorted according to a descending sorting manner based on the number of the subsequences in each category; selecting target classes with the quantity smaller than a set threshold value from the sorting result; and mapping the subsequences in the target class to the coding sequence of the audio data based on the mode of shifting and dividing the coding sequence to obtain a noise distribution region, namely the noise distribution region in the audio data.

S503, denoising the noise distribution area.

Specifically, after the noise distribution region is determined, filtering processing is performed on the noise distribution region. Optionally, denoising processing may be performed by a method of eliminating the subsequence, so as to implement denoising processing on the audio data.

In the embodiment, the sub-sequences are clustered according to the statistical characteristics to determine the noise distribution area in the audio data, and compared with a related audio denoising method, the noise characteristics do not need to be labeled in advance, so that the complexity of audio denoising is reduced, and the use scene of audio denoising processing is widened. And meanwhile, denoising is carried out on the noise distribution area, and the accuracy of audio denoising is improved.

Fig. 6 is a schematic flow chart of performing an audio denoising process based on statistical characteristics in another embodiment, and based on the above embodiment, this embodiment further explains S204 in detail, which specifically includes the following steps:

s601, according to the first probability distribution in the statistical characteristics of the subsequences, determining the first probability distribution of the shift sequence concentrated shift sequence.

Specifically, for each shift sequence in the shift sequence set, the number of subsequences having the same first probability distribution is counted based on the first probability distribution in the statistical characteristics of the subsequences in the subsequence set corresponding to the shift sequence, and then the first probability distribution of the shift sequence is determined.

For example, if there are 100 subsequences with P =0.2, 110 subsequences with P =0.3, and 200 subsequences with P =0.5 in the set of subsequences corresponding to the shifted sequence, then the first probability distribution of the shifted sequence at this time can be represented as: { (0.2,100), (0.3,110), (0.5,200) }.

S602, a first probability distribution of the shift sequence set is obtained according to the first probability distribution of the shift sequences.

Specifically, the first probability distribution of each shift sequence in the shift sequence set is statistically analyzed, so that the first probability distribution of the entire shift sequence set can be obtained. For example, the same P in each shifted sequence can be linearly added to obtain a first probability distribution for the entire set of shifted sequences.

For example, the shifted sequence set includes 3 shifted sequences, wherein the first probability distribution of the shifted sequence 1 is { (0.2,100), (0.3,110), (0.5,200) }, the probability distribution of the shifted sequence 2 is { (0.1,100), (0.4,110), (0.5,200) }, and the probability distribution of the shifted sequence 3 is { (0.6,100), (0.3,110), (0.1,200) }, and the shifted sequences are linearly added, and the first probability distribution of the entire shifted sequence set is { (0.1,300), (0.2,100), (0.3,220), (0.4,110), (0.5,400), (0.6,100) }.

S603, determining a second probability distribution of the shift bit sequence in the shift sequence set according to the second probability distribution in the statistical characteristics of the subsequences.

Corresponding to step S601, for each shifted sequence in the shifted sequence set, based on the second probability distribution in the statistical characteristics of each subsequence in the subsequence set corresponding to the shifted sequence, the number of subsequences having the same second probability distribution is counted, and then the second probability distribution of the shifted sequence is determined.

For example, if there are 120 sub-sequences with Q =0.15, 150 sub-sequences with Q =0.2, and 180 sub-sequences with Q =0.33 in the sub-sequence set corresponding to the shifted sequence, then the second probability distribution of the shifted sequence may be represented as: { (0.15,120), (0.2,150), (0.33,180) }.

S604, according to the second probability distribution of the shift sequences, obtaining a second probability distribution of the shift sequence set.

Similarly, the second probability distribution of each shift sequence in the shift sequence set is statistically analyzed, and the second probability distribution of the entire shift sequence set can be obtained. For example, the same Q in each shifted sequence can be linearly added to obtain a second probability distribution for the entire set of shifted sequences.

For example, the shifted sequence set includes 3 shifted sequences, wherein the second probability distribution of the shifted sequence 1 is { (0.15,120), (0.2,150), (0.33,180) }, the probability distribution of the shifted sequence 2 is { (0.1,100), (0.33,110), (0.5,200) }, the probability distribution of the shifted sequence 3 is { (0.33,100), (0.5,110), (0.7,200) }, and linear additions are performed for the shifted sequences, and the first probability distribution of the entire shifted sequence set is { (0.1,100), (0.15,120), (0.2,150), (0.33,220), (0.4,390), (0.5,310), (0.7,200) }.

And S605, denoising the audio data according to the first probability distribution and the second probability distribution of the shift sequence set.

Specifically, a first probability distribution P and a second probability distribution Q in a shift sequence set are combined, then the obtained probability combination of P and Q is compared with the probability combination of the subsequence, and if the probability combination exists in the subsequence, the probability combination is effective; if the probability combination does not exist in the subsequence, the probability combination is invalid.

Further, based on the valid P, Q probability combinations, a series of (X, Y, Z) coordinates are obtained. Where X represents the first probability distribution, Y represents the second probability distribution, and Z represents the number of occurrences of the probability combination, such as a probability combination of P =0.2, Q =0.5, where Z = C1 (i.e., the number of occurrences of P =0.2 in the shifted sequence set) + C2 (i.e., the number of occurrences of Q =0.5 in the shifted sequence set). And then, constructing a three-dimensional stereo map based on a series of (X, Y, Z) coordinates obtained by conversion, wherein probability combinations with fewer occurrences in the subsequence can be visually obtained from the map, and the probability combinations with the occurrences less than a preset threshold value can be regarded as the statistical characteristics of the noise.

Furthermore, based on the manner of shifting and dividing the coding sequence, the probability combination regarded as the noise statistical characteristic is mapped to the coding sequence of the audio data, so as to obtain a noise distribution region, i.e. a noise distribution region in the audio data. And after the noise distribution area is determined, denoising is carried out on the noise distribution area.

In this embodiment, based on the logic of shifting and dividing the coding sequence, the first probability distribution and the second probability distribution of the shifted sequence set can be derived according to the statistical characteristics of the subsequences, and then based on the probability distribution of the shifted sequence set, the denoising processing of the audio data can be realized.

Fig. 7 is a flowchart illustrating an audio denoising method in another embodiment, and on the basis of the foregoing embodiment, this embodiment provides an alternative example of the audio denoising method. With reference to fig. 7, the specific implementation process is as follows:

s701, carrying out shift transformation on the coding sequence of the audio data to obtain a shift sequence set.

S702, based on the first set length, dividing the shift sequences in the shift sequence set to obtain a window sequence set corresponding to the shift sequences.

And S703, dividing the window sequences in the window sequence set corresponding to the shift sequence based on the second set length to obtain a subsequence set corresponding to the shift sequence.

S704, aiming at each subsequence in the subsequence set, determining the statistical characteristics of the subsequence according to the length of the subsequence, the number of first target character strings in the subsequence and the number of second target character strings in the subsequence.

Wherein the first target string is different from the target string.

S705, a first probability distribution of the sequence of shifts in the set of shifted sequences is determined according to the first probability distribution in the statistical characteristics of the subsequences.

S706, according to the first probability distribution of the shift sequences, a first probability distribution of the shift sequence set is obtained.

And S707, determining a second probability distribution of the shift sequence in the shift sequence set according to the second probability distribution in the statistical characteristics of the subsequences.

S708, obtaining a second probability distribution of the shifted sequence set according to the second probability distribution of the shifted sequences.

And S709, denoising the audio data according to the first probability distribution and the second probability distribution of the displacement sequence set.

For the specific processes of S701 to S709, reference may be made to the description of the method embodiments, which implement the principle and the technical effect similar to each other, and further description is omitted here.

It should be understood that, although the steps in the flowcharts related to the embodiments as described above are sequentially displayed as indicated by arrows, the steps are not necessarily performed sequentially as indicated by the arrows. The steps are not performed in the exact order shown and described, and may be performed in other orders, unless explicitly stated otherwise. Moreover, at least a part of the steps in the flowcharts related to the embodiments described above may include multiple steps or multiple stages, which are not necessarily performed at the same time, but may be performed at different times, and the execution order of the steps or stages is not necessarily sequential, but may be rotated or alternated with other steps or at least a part of the steps or stages in other steps.

Based on the same inventive concept, the embodiment of the application also provides an audio denoising device for realizing the audio denoising method. The implementation scheme for solving the problem provided by the apparatus is similar to the implementation scheme described in the above method, so the specific limitations in one or more embodiments of the audio denoising apparatus provided below can be referred to the limitations of the audio denoising method in the above, and are not described herein again.

In one embodiment, as shown in fig. 8, there is provided an audio denoising apparatus 1, including: data shift module 10, sequence division module 20, measure analysis module 30 and audio denoising module 40, wherein:

the data shifting module 10 is configured to perform shifting transformation on a coding sequence of the audio data to obtain a shifting sequence set;

a sequence dividing module 20, configured to divide the shift sequences in the shift sequence set to obtain a subsequence set corresponding to the shift sequences;

the measure analysis module 30 is configured to perform measure analysis on the target character string for the subsequences in the subsequence set to obtain statistical characteristics of the subsequences;

and the audio denoising module 40 is configured to perform denoising processing on the audio data according to the statistical characteristics of the subsequences.

In one embodiment, as shown in fig. 9, the sequence dividing module 20 in fig. 8 further includes:

the first dividing unit 21 is configured to divide the shift sequences in the shift sequence set based on a first set length to obtain a window sequence set corresponding to the shift sequences;

a second dividing unit 22, configured to divide the window sequences in the window sequence set corresponding to the shift sequence based on a second set length to obtain a sub-sequence set corresponding to the shift sequence; wherein the first set length is greater than the second set length.

In one embodiment, the metric analysis module 30 is configured to:

for each subsequence in the set of subsequences, determining a statistical characteristic of the subsequence according to the length of the subsequence, the number of first target character strings in the subsequence and the number of second target character strings in the subsequence; wherein the first target string is different from the target string.

In one embodiment, as shown in fig. 10, the measure analyzing module 30 in fig. 8 further includes:

a first analyzing unit 31, configured to use a ratio between the length of the subsequence and the number of first target strings in the subsequence as a first probability distribution in the statistical features of the subsequence;

a second analyzing unit 32, configured to use a ratio between the length of the subsequence and the number of second target strings in the subsequence as a second probability distribution in the statistical characteristics of the subsequence.

In one embodiment, the audio denoising module 40 is configured to:

clustering the subsequences according to the statistical characteristics;

and denoising the noise distribution area.

In one embodiment, the audio denoising module 40 is further configured to:

determining a first probability distribution of the shift sequences in the set of shift sequences according to the first probability distribution in the statistical features of the subsequences;

and denoising the audio data according to the first probability distribution and the second probability distribution of the shift sequence set.

The modules in the audio denoising apparatus can be wholly or partially implemented by software, hardware and a combination thereof. The modules can be embedded in a hardware form or independent from a processor in the computer device, and can also be stored in a memory in the computer device in a software form, so that the processor can call and execute operations corresponding to the modules.

In one embodiment, a computer device is provided, which may be a server, and its internal structure diagram may be as shown in fig. 11. The computer device includes a processor, a memory, and a network interface connected by a system bus. Wherein the processor of the computer device is configured 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, a computer program, and a database. The internal memory provides an environment for the operation of an operating system and computer programs in the non-volatile storage medium. The database of the computer device is used for storing audio data to be denoised and the like. The network interface of the computer device is used for communicating with an external terminal through a network connection. The computer program is executed by a processor to implement an audio denoising method.

Those skilled in the art will appreciate that the architecture shown in fig. 11 is merely a block diagram of some of the structures associated with the disclosed aspects and is not intended to limit the computing devices to which the disclosed aspects apply, as particular computing devices may include more or less components than those shown, or may combine certain components, or have a different arrangement of components.

In one embodiment, a computer device is provided, comprising a memory having a computer program stored therein and a processor that when executing the computer program performs the steps of:

On the basis of the above embodiment, when the processor executes the logic of dividing the shift sequences in the shift sequence set in the computer program to obtain the sub-sequence sets corresponding to the shift sequences, the following steps are specifically implemented:

dividing the shift sequences in the shift sequence set based on the first set length to obtain a window sequence set corresponding to the shift sequences; based on a second set length, dividing the window sequences in the window sequence set corresponding to the shift sequences to obtain a subsequence set corresponding to the shift sequences; wherein the first set length is greater than the second set length.

On the basis of the above embodiment, when the processor executes the logic for performing the measure analysis of the target character string on the subsequences in the subsequence set in the computer program to obtain the statistical characteristics of the subsequences, the following steps are specifically implemented:

On the basis of the above embodiment, when the processor executes the logic in the computer program for determining the statistical characteristics of the subsequence according to the length of the subsequence, the number of first target character strings in the subsequence, and the number of second target character strings in the subsequence, the following steps are specifically implemented:

taking the ratio of the length of the subsequence to the number of first target character strings in the subsequence as a first probability distribution in the statistical characteristics of the subsequence; and taking the ratio of the length of the subsequence to the number of second target character strings in the subsequence as a second probability distribution in the statistical characteristics of the subsequence.

On the basis of the above embodiment, when the processor executes the logic for performing denoising processing on the audio data according to the statistical characteristics of the subsequence in the computer program, the following steps are specifically implemented:

clustering the subsequences according to the statistical characteristics; determining a noise distribution area in the audio data according to the number of the subsequences in each type; and denoising the noise distribution area.

determining a first probability distribution of the sequence of shifts in the set of shift sequences based on the first probability distribution in the statistical features of the subsequences; obtaining a first probability distribution of the shift sequence set according to the first probability distribution of the shift sequences; determining a second probability distribution of the shift sequence in the shift sequence set according to the second probability distribution in the statistical characteristics of the subsequences; obtaining a second probability distribution of the shift sequence set according to the second probability distribution of the shift sequence; and denoising the audio data according to the first probability distribution and the second probability distribution of the displacement sequence set.

In one embodiment, a computer-readable storage medium is provided, having a computer program stored thereon, which when executed by a processor, performs the steps of:

On the basis of the above embodiment, when the code logic that divides the shift sequences in the shift sequence set in the computer program to obtain the sub-sequence set corresponding to the shift sequences is executed by the processor, the following steps are specifically implemented:

dividing the shift sequences in the shift sequence set based on the first set length to obtain a window sequence set corresponding to the shift sequences; based on the second set length, dividing the window sequences in the window sequence set corresponding to the shift sequences to obtain a subsequence set corresponding to the shift sequences; wherein the first set length is greater than the second set length.

On the basis of the above embodiment, when the code logic for performing measure analysis of the target character string on the sub-sequences in the sub-sequence set in the computer program and obtaining the statistical characteristics of the sub-sequences is executed by the processor, the following steps are specifically implemented:

On the basis of the above embodiment, when the code logic in the computer program that determines the statistical characteristics of the subsequence according to the length of the subsequence, the number of first target strings in the subsequence and the number of second target strings in the subsequence is executed by the processor, the following steps are specifically implemented:

On the basis of the above embodiment, when the code logic for denoising the audio data according to the statistical characteristics of the subsequence in the computer program is executed by the processor, the following steps are specifically implemented:

clustering the subsequences according to the statistical characteristics; determining a noise distribution area in the audio data according to the number of the various subsequences; and denoising the noise distribution area.

determining a first probability distribution of the sequence of shifts in the set of shift sequences based on the first probability distribution in the statistical features of the subsequences; obtaining a first probability distribution of the shift sequence set according to the first probability distribution of the shift sequences; determining a second probability distribution of the shift sequences in the shift sequence set according to a second probability distribution in the statistical characteristics of the subsequences; obtaining a second probability distribution of the shift sequence set according to the second probability distribution of the shift sequence; and denoising the audio data according to the first probability distribution and the second probability distribution of the shift sequence set.

In one embodiment, a computer program product is provided, comprising a computer program which when executed by a processor performs the steps of:

On the basis of the above embodiment, when the computer program is executed by the processor to perform the operation of dividing the shift sequences in the shift sequence set to obtain the sub-sequence sets corresponding to the shift sequences, the following steps are specifically implemented:

dividing the shift sequences in the shift sequence set based on the first set length to obtain a window sequence set corresponding to the shift sequences; based on a second set length, dividing the window sequences in the window sequence set corresponding to the shift sequences to obtain a subsequence set corresponding to the shift sequences; wherein, the first set length is greater than the second set length.

On the basis of the above embodiment, when the computer program is executed by the processor to perform the measurement analysis of the target character string on the sub-sequences in the sub-sequence set to obtain the statistical characteristics of the sub-sequences, the following steps are specifically implemented:

On the basis of the above embodiment, when the computer program is executed by the processor to determine the statistical characteristics of the subsequence according to the length of the subsequence, the number of first target character strings in the subsequence, and the number of second target character strings in the subsequence, the following steps are specifically implemented:

On the basis of the above embodiment, when the computer program is executed by the processor to perform the operation of denoising the audio data according to the statistical characteristics of the subsequence, the following steps are specifically implemented:

determining a first probability distribution of the shift sequences in the set of shift sequences according to the first probability distribution in the statistical features of the subsequences; obtaining a first probability distribution of the shift sequence set according to the first probability distribution of the shift sequences; determining a second probability distribution of the shift sequence in the shift sequence set according to the second probability distribution in the statistical characteristics of the subsequences; obtaining a second probability distribution of the shift sequence set according to the second probability distribution of the shift sequence; and denoising the audio data according to the first probability distribution and the second probability distribution of the shift sequence set.

It should be noted that, the audio information (including but not limited to the audio information to be denoised, etc.) and the data (including but not limited to the data for analysis, the stored data, the displayed data, etc.) referred to in the present application are both information and data authorized by the user or sufficiently authorized by each party.

It will be understood by those skilled in the art that all or part of the processes of the methods of the embodiments described above can be implemented by hardware instructions of a computer program, which can be stored in a non-volatile computer-readable storage medium, and when executed, can include the processes of the embodiments of the methods described above. Any reference to memory, database, or other medium used in the embodiments provided herein may include at least one of non-volatile and volatile memory. The nonvolatile Memory may include Read-Only Memory (ROM), magnetic tape, floppy disk, flash Memory, optical Memory, high-density embedded nonvolatile Memory, resistive Random Access Memory (ReRAM), magnetic Random Access Memory (MRAM), ferroelectric Random Access Memory (FRAM), phase Change Memory (PCM), graphene Memory, and the like. Volatile Memory can include Random Access Memory (RAM), external cache Memory, and the like. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM), among others. The databases referred to in various embodiments provided herein may include at least one of relational and non-relational databases. The non-relational database may include, but is not limited to, a block chain based distributed database, and the like. The processors referred to in the embodiments provided herein may be general purpose processors, central processing units, graphics processors, digital signal processors, programmable logic devices, quantum computing based data processing logic devices, etc., without limitation.

The technical features of the above embodiments can be arbitrarily combined, and for the sake of brevity, all possible combinations of the technical features in the above embodiments are not described, but should be considered as the scope of the present specification as long as there is no contradiction between the combinations of the technical features.

The above-mentioned embodiments only express several embodiments of the present application, and the description thereof is more specific and detailed, but not construed as limiting the scope of the present application. It should be noted that, for a person skilled in the art, several variations and modifications can be made without departing from the concept of the present application, which falls within the scope of protection of the present application. Therefore, the protection scope of the present application shall be subject to the appended claims.

Claims

1. A method for denoising audio, the method comprising:

2. The method according to claim 1, wherein the dividing the shifted sequences in the set of shifted sequences to obtain a set of sub-sequences corresponding to the shifted sequences comprises:

dividing the shift sequences in the shift sequence set based on a first set length to obtain a window sequence set corresponding to the shift sequences;

based on a second set length, dividing the window sequences in the window sequence set corresponding to the shift sequences to obtain a subsequence set corresponding to the shift sequences; wherein the first set length is greater than the second set length.

3. The method according to claim 1 or 2, wherein the performing a measure analysis on the subsequences in the set of subsequences to obtain statistical characteristics of the subsequences comprises:

4. The method of claim 3, wherein determining the statistical characteristics of the subsequence based on the length of the subsequence, the number of first target strings in the subsequence, and the number of second target strings in the subsequence comprises:

5. The method according to claim 4, wherein said denoising the audio data according to the statistical characteristics of the sub-sequence comprises:

clustering the subsequences according to the statistical characteristics;

determining a noise distribution area in the audio data according to the number of the subsequences of each type;

and denoising the noise distribution region.

6. The method according to claim 4, wherein said denoising the audio data according to the statistical characteristics of the sub-sequence comprises:

determining a first probability distribution of the sequence of shifts in the set of shifted sequences according to the first probability distribution in the statistical features of the subsequences;

determining a second probability distribution of the shift sequences in the shift sequence set according to a second probability distribution in the statistical characteristics of the subsequences;

7. An apparatus for audio denoising, the apparatus comprising:

the data shifting module is used for shifting and transforming the coding sequence of the audio data to obtain a shifting sequence set;

and the audio denoising module is used for denoising the audio data according to the statistical characteristics of the subsequence.

8. A computer device comprising a memory and a processor, the memory storing a computer program, characterized in that the processor, when executing the computer program, implements the steps of the method of any of claims 1 to 6.

9. A computer-readable storage medium, on which a computer program is stored, which, when being executed by a processor, carries out the steps of the method of any one of claims 1 to 6.

10. A computer program product comprising a computer program, characterized in that the computer program realizes the steps of the method of any one of claims 1 to 6 when executed by a processor.