A method and device for noise recognition and a computer-readable storage medium

Abstract

The present invention discloses a noise recognition method, apparatus, and computer-readable storage medium. The method includes respectively collecting audio in an area only affected by road traffic noise and an area only affected by social life noise to obtain sample audio data; performing signal processing and feature extraction on the sample audio data to obtain a training dataset; training a deep convolutional neural network model according to the training dataset to obtain a sound source recognition model; and recognizing the noise in the area to be recognized according to the sound source recognition model to determine the type of the noise sound source in the area to be recognized. The present invention can distinguish the main noise sources in urban areas, thereby identifying the areas mainly affected by road traffic noise, helping to effectively prevent and control traffic noise pollution, and can be widely applied to the field of computer technology.

Description

Technical Field

[0001] The present invention relates to the field of computer technology, and in particular, to a noise recognition method, device, and computer-readable storage medium. Background Art

[0002] The urban sound environment is the result of the combined action of multiple sound sources. Among them, the distribution ranges of social life noise and traffic noise are much larger than those of industrial production noise and construction noise. Different from the characteristics of relatively small numbers, easy to enclose, and controllable working hours in industrial production areas and construction areas, the individual sound sources of social life noise and traffic noise are numerous, and it is often difficult to separate them from the surrounding environment through physical facilities, and the influence duration is long. Therefore, in most urban areas, traffic noise and social life noise are the main noise pollution sources. The source of road traffic noise is urban roads, which has the characteristics of fixed sound source areas and easy control compared with social life noise. Therefore, in the current urban sound environment governance, the prevention and control of traffic noise is the main focus.

[0003] In the current noise pollution assessments at home and abroad, the main indicators for evaluating the degree of noise pollution are the equivalent sound level of the noise and the magnitude exceeding the environmental noise limit. At present, traffic noise is the main prevention and control target, and the measures for treating noise pollution mainly include building sound barriers, speed limits, traffic flow limits, laying noise-reducing road surfaces, etc., which have little impact on social life noise. Judging the main noise pollution sources according to the noise spectrum characteristics of the sound sources has a guiding role in the current urban environmental noise pollution governance.

[0004] For some areas of the city, the main noise pollution sources may be traffic noise or social life noise. Therefore, it is necessary to form a method for classifying and discriminating sound sources based on noise data to distinguish road traffic noise and social life noise. Summary of the Invention

[0005] In view of this, embodiments of the present invention provide a noise recognition method, device, and computer-readable storage medium to distinguish the main noise sources in urban areas, thereby identifying the areas mainly affected by road traffic noise and helping to carry out effective prevention and control of traffic noise pollution.

[0006] One aspect of an embodiment of the present invention provides a noise recognition method, including:

[0007] Performing audio collection from areas affected only by road traffic noise and areas affected only by social life noise respectively to obtain sample audio data;

[0008] Performing signal processing and feature extraction on the sample audio data to obtain a training dataset to be trained;

[0009] Train a deep convolutional neural network model based on the training dataset to obtain a sound source recognition model;

[0010] Identify the noise in the area to be identified according to the sound source recognition model, and determine the type of the noise source in the area to be identified.

[0011] Optionally, the step of separately collecting audio samples from areas affected only by road traffic noise and areas affected only by social life noise to obtain sample audio data includes:

[0012] Select areas affected only by road traffic noise and areas affected only by social life noise, collect environmental noise and record audio to obtain sample audio signals;

[0013] Perform data augmentation operations on the sample audio signals to obtain sample audio data;

[0014] Among them, the data augmentation operations include operations of adding Gaussian white noise, waveform displacement, and waveform stretching.

[0015] Optionally, the step of performing signal processing and feature extraction on the sample audio data to obtain a training dataset includes:

[0016] Segment the sample audio data to obtain multiple audio segments with a time length of 5 seconds;

[0017] Perform short-time Fourier transform on the segmented audio segments to obtain the spectra of the audio segments;

[0018] Construct a Mel filter, filter the spectra of the audio segments to obtain Mel log spectrograms;

[0019] After performing discrete cosine transform on the Mel log spectrograms, select Mel cepstral coefficients to obtain MFCC features;

[0020] Render the Mel log spectrograms and the MFCC features into graphs to obtain Mel log spectrogram graphs and MFCC feature graphs of road traffic noise and social life noise.

[0021] Optionally, the step of training a deep convolutional neural network model based on the training dataset to obtain a sound source recognition model includes:

[0022] Construct a deep convolutional neural network model by stacking two-dimensional convolutional layers and max pooling layers;

[0023] Use the Mel log spectrogram graphs and MFCC feature graphs of road traffic noise and social life noise as inputs respectively to train the model.

[0024] Optionally, identifying the noise in the area to be identified according to the sound source recognition model to determine the type of the noise sound source in the area to be identified includes:

[0025] Collecting audio data of the ambient noise in the area to be identified, filtering the audio data, and generating a Mel logarithmic spectrogram and an MFCC feature map;

[0026] Inputting the Mel logarithmic spectrogram and the MFCC feature map into the trained sound source recognition model, identifying the features of the spectrogram and the MFCC feature map, and further discriminating the sound source category according to the identified features;

[0027] Judging the main ambient noise sound source in the area to be identified according to the comparison of the recognition quantities of road traffic noise and social life noise in the sound source category discrimination result.

[0028] Optionally, the signal processing and feature extraction of the sample audio data to obtain the dataset to be trained further includes:

[0029] Determining the Hertz frequency range of audio signal processing according to the audible threshold range of the human ear;

[0030] Determining the start and end frequencies of the Mel filter according to the correspondence between the Mel frequency and the Hertz frequency;

[0031] Constructing a Mel filter with 100 center frequencies, filtering the audio signal, converting the audio signal with continuous frequencies into discrete audio signals at 100 Mel center frequencies, and obtaining a Mel logarithmic spectrum;

[0032] Performing discrete cosine transform on the Mel logarithmic spectrum, and taking the 2nd to 13th coefficients after the transform as Mel cepstral coefficients;

[0033] Drawing the spectrogram of the Mel logarithmic spectrum and the MFCC coefficient feature map, so as to obtain the Mel logarithmic spectrogram and the MFCC feature map respectively representing road traffic noise and social life noise.

[0034] Another aspect of the embodiment of the present invention further provides a noise recognition device, including:

[0035] A first module, configured to collect audio respectively from an area affected only by road traffic noise and an area affected only by social life noise to obtain sample audio data;

[0036] A second module, configured to perform signal processing and feature extraction on the sample audio data to obtain a dataset to be trained;

[0037] A third module, configured to train a deep convolutional neural network model according to the dataset to be trained to obtain a sound source recognition model;

[0038] The fourth module is configured to identify the noise in the area to be identified according to the sound source recognition model, and determine the type of the noise sound source in the area to be identified.

[0039] Another aspect of the embodiments of the present invention further provides an electronic device, including a processor and a memory;

[0040] The memory is used to store programs;

[0041] The processor executes the program to implement the method as described above.

[0042] Another aspect of the embodiments of the present invention further provides a computer-readable storage medium, where the storage medium stores a program, and the program is executed by a processor to implement the method as described above.

[0043] The embodiments of the present invention also disclose a computer program product or a computer program. The computer program product or the computer program includes computer instructions, and the computer instructions are stored in a computer-readable storage medium. The processor of the computer device can read the computer instructions from the computer-readable storage medium, and the processor executes the computer instructions, so that the computer device executes the method as described above.

[0044] The embodiments of the present invention respectively perform audio collection from areas affected only by road traffic noise and areas affected only by social life noise to obtain sample audio data; perform signal processing and feature extraction on the sample audio data to obtain a dataset to be trained; train a deep convolutional neural network model according to the dataset to be trained to obtain a sound source recognition model; identify the noise in the area to be identified according to the sound source recognition model, and determine the type of the noise sound source in the area to be identified. The present invention can distinguish the main noise sources in urban areas, thereby identifying the areas mainly affected by road traffic noise, and helping to effectively prevent and control traffic noise pollution. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the following will briefly introduce the drawings required for the description of the embodiments. Obviously, the following drawings are only some embodiments of the present application. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0046] Figure 1 It is a schematic flowchart of the method for identifying noise sources of the present invention;

[0047] Figure 2 It is the Mel logarithmic spectrogram of road traffic noise;

[0048] Figure 3It is the MFCC feature map of road traffic noise;

[0049] Figure 4 It is the mel logarithmic spectrogram of social life noise;

[0050] Figure 5 It is the MFCC feature map of social life noise. Specific implementation manners

[0051] In order to make the purpose, technical solutions and advantages of the present application clearer, the present application will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0052] Aiming at the problems existing in the prior art, the purpose of the present invention is to propose a method for identifying road traffic noise and social life noise in urban areas, which is used to distinguish the main noise sources in urban areas, so as to identify the areas mainly affected by road traffic noise and help to carry out effective prevention and control of traffic noise pollution.

[0053] Specifically, one aspect of the embodiments of the present invention provides a noise identification method, including:

[0054] Audio is respectively collected from areas only affected by road traffic noise and areas only affected by social life noise to obtain sample audio data;

[0055] The sample audio data is subjected to signal processing and feature extraction to obtain a training dataset to be trained;

[0056] The deep convolutional neural network model is trained according to the training dataset to be trained to obtain a sound source identification model;

[0057] The noise of the area to be identified is identified according to the sound source identification model to determine the type of the noise sound source in the area to be identified.

[0058] Optionally, the audio is respectively collected from areas only affected by road traffic noise and areas only affected by social life noise to obtain sample audio data, including:

[0059] Areas only affected by road traffic noise and areas only affected by social life noise are selected to collect environmental noise and record audio to obtain sample audio signals;

[0060] The sample audio signals are subjected to data augmentation operations to obtain sample audio data;

[0061] Among them, the data augmentation operations include operations of adding Gaussian white noise, waveform displacement and waveform stretching.

[0062] Optionally, the signal processing and feature extraction of the sample audio data to obtain a training dataset includes:

[0063] Segment the sample audio data to obtain multiple audio segments with a time length of 5 seconds;

[0064] Perform short-time Fourier transform on the segmented audio segments to obtain the spectra of each audio segment;

[0065] Construct a Mel filter, filter the spectrum of the audio segment to obtain a Mel log spectrum;

[0066] After performing discrete cosine transform on the Mel log spectrum, select Mel cepstral coefficients to obtain MFCC features;

[0067] Render the Mel log spectrum and the MFCC features into graphs to obtain the Mel log spectrogram and MFCC feature map of road traffic noise and social life noise.

[0068] Optionally, the training of the deep convolutional neural network model according to the training dataset to obtain a sound source recognition model includes:

[0069] Construct a deep convolutional neural network model by stacking two-dimensional convolutional layers and max-pooling layers;

[0070] Use the Mel log spectrogram and MFCC feature map of road traffic noise and social life noise as inputs respectively to train the model.

[0071] Optionally, the identification of the noise in the area to be identified according to the sound source recognition model to determine the type of sound source of the noise in the area to be identified includes:

[0072] Collect the audio data of the environmental noise in the area to be identified, filter the audio data to generate a Mel log spectrogram and an MFCC feature map;

[0073] Input the Mel log spectrogram and MFCC feature map into the trained sound source recognition model, identify the features of the spectrogram and MFCC feature map, and then discriminate the sound source category of the identified features;

[0074] According to the comparison of the recognition quantities of road traffic noise and social life noise in the sound source category discrimination result, judge the main environmental noise source in the area to be identified.

[0075] Optionally, the signal processing and feature extraction of the sample audio data to obtain a training dataset further includes:

[0076] Determine the Hertz frequency range of audio signal processing according to the audible threshold range of the human ear;

[0077] Determine the start and end frequencies of the Mel filter according to the correspondence between Mel frequency and Hertz frequency;

[0078] Construct a Mel filter with 100 center frequencies, filter the audio signal, convert the audio signal with continuous frequencies into discrete audio signals at 100 Mel center frequencies, and obtain the Mel log spectrum;

[0079] Perform discrete cosine transform on the Mel log spectrum, and take the 2nd to 13th coefficients after the transform as the Mel cepstral coefficients;

[0080] Draw the spectrogram of the Mel log spectrum and the MFCC coefficient feature map, so as to obtain the Mel log spectrogram and MFCC feature map representing road traffic noise and social life noise respectively.

[0081] Another aspect of the embodiments of the present invention also provides a noise recognition device, including:

[0082] The first module is used to collect audio respectively from areas only affected by road traffic noise and areas only affected by social life noise to obtain sample audio data;

[0083] The second module is used to perform signal processing and feature extraction on the sample audio data to obtain a training dataset;

[0084] The third module is used to train a deep convolutional neural network model according to the training dataset to obtain a sound source recognition model;

[0085] The fourth module is used to recognize the noise in the area to be recognized according to the sound source recognition model to determine the type of the noise sound source in the area to be recognized.

[0086] Another aspect of the embodiments of the present invention also provides an electronic device, including a processor and a memory;

[0087] The memory is used to store programs;

[0088] The processor executes the program to implement the method described above.

[0089] Another aspect of the embodiments of the present invention also provides a computer-readable storage medium, where the storage medium stores a program, and the program is executed by a processor to implement the method described above.

[0090] An embodiment of the present invention also discloses a computer program product or a computer program. The computer program product or the computer program includes computer instructions, and the computer instructions are stored in a computer-readable storage medium. A processor of a computer device can read the computer instructions from the computer-readable storage medium, and the processor executes the computer instructions, so that the computer device executes the foregoing method.

[0091] The following will describe the specific implementation process of the present invention in detail with reference to the accompanying drawings of the specification:

[0092] A method for identifying road traffic noise and social life noise in urban areas, as Figure 1 shown, the method includes the following steps:

[0093] S1. Audio acquisition of road traffic noise and social life noise. Select areas affected only by road traffic noise and areas affected only by social life noise, collect environmental noise, record audio, and perform data enhancement operations on the recorded audio, including adding Gaussian white noise, waveform displacement, and waveform stretching.

[0094] S2. Audio signal filtering and feature extraction. Segment the audio signal to obtain multiple audio segments with a time length of 5 seconds; perform short-time Fourier transform on the segmented audio segments to obtain the spectra of the audio segments; construct a Mel filter bank, filter the spectra of the audio segments to obtain Mel log spectrograms; perform discrete cosine transform on the Mel log spectrograms, and take the 2nd to 13th coefficients as Mel cepstral coefficients, that is, MFCC features. Render the Mel log spectrograms and MFCC features into graphs to obtain the Mel log spectrogram and MFCC feature graph of road traffic noise and social life noise.

[0095] S3. Training of the deep convolutional neural network model. Stack two-dimensional convolutional layers and max-pooling layers to construct a deep convolutional neural network model, and use the Mel log spectrogram and MFCC feature graph of road traffic noise and social life noise as inputs respectively to train the model to achieve the judgment of the sound source category.

[0096] S4. Identification of the main environmental noise sources in the city. Collect audio data of environmental noise in the research area, filter the audio data to generate a Mel log spectrogram and an MFCC feature graph, input them into the trained deep convolutional neural network model, identify the main features of the spectrogram and the MFCC feature graph, and determine which sound source category it belongs to. By comparing the number of identifications of road traffic noise and social life noise, judge whether the main environmental noise source in the research area is road traffic noise.

[0097] For the collection of road traffic noise and social life noise, areas within the city that are only affected by road traffic noise should be selected to collect road traffic noise audio, and areas that are not affected by road traffic noise, construction noise, and industrial production noise should be selected to collect social life noise audio.

[0098] For the filtering of the audio signal, first, according to the audible threshold range of the human ear, the Hertz frequency range for audio signal processing is determined to be 10 - 20000 Hz. According to the corresponding relationship between Mel frequency and Hertz frequency the start and end frequencies of the Mel filter are determined; a Mel filter with 100 center frequencies is constructed to filter the audio signal, converting the audio signal with continuous frequencies into discrete audio signals at 100 Mel center frequencies to obtain the Mel logarithmic spectrogram; the Mel logarithmic spectrogram is subjected to discrete cosine transform, and the 2nd to 13th coefficients after the transform are taken as the Mel cepstral coefficients (MFCC coefficients); the spectrogram of the Mel logarithmic spectrogram and the MFCC coefficient feature map are plotted, thereby obtaining the Mel logarithmic spectrogram and MFCC feature map respectively representing road traffic noise and social life noise, where Figure 2 is the Mel logarithmic spectrogram of road traffic noise; Figure 3 is the MFCC feature map of road traffic noise; Figure 4 is the Mel logarithmic spectrogram of social life noise; Figure 5 is the MFCC feature map of social life noise.

[0099] For the training of the deep convolutional neural network model, two deep convolutional neural network models each containing 4 two-dimensional convolutional layers and 4 max-pooling layers are constructed by alternately stacking the two-dimensional convolutional layer and the max-pooling layer; the Mel logarithmic spectrogram of road traffic noise and social life noise, and the MFCC feature map of road traffic noise and social life noise are respectively used as the input of the model, and two deep convolutional neural network models are trained to respectively perform feature recognition on the Mel logarithmic spectrogram and the MFCC feature map, thereby realizing sound source classification.

[0100] For the identification of the main environmental noise in the city, environmental noise audio data within the research area is collected, and the audio data is segmented every 5 s; the segmented audio segments are subjected to Mel filtering to obtain the Mel logarithmic spectrogram and the MFCC feature map; the trained deep convolutional neural network model is used to classify the Mel logarithmic spectrogram and the MFCC feature map to obtain the classification results of each audio segment; the proportions of the Mel logarithmic spectrogram and the MCFF feature map respectively judged as road traffic noise and social life noise are statistically analyzed, and the main sound source category in the environmental noise is determined according to this proportion.

[0101] Compared with the prior art, the present invention has the following advantages:

[0102] The most important technical breakthrough of the present invention lies in being able to identify the main noise sources in environmental noise, thereby screening out the areas mainly affected by road traffic noise, so as to facilitate the control of noise pollution. The present invention fully considers the characteristics of the mixed influence of various sound sources on the target area in the city. By the spectral characteristics of sound, according to a large number of audio samples, a deep convolutional neural network model is trained to classify the characteristics of the environmental noise in the research area, which helps to identify the main noise source categories in each time segment. By analyzing the proportion of different noise source categories, the category of the main environmental noise source is determined, so as to clarify the key areas for the treatment of urban road traffic noise pollution.

[0103] In some alternative embodiments, the functions / operations mentioned in the block diagrams may not occur in the order mentioned in the operation diagrams. For example, depending on the functions / operations involved, two consecutive blocks shown may actually be executed substantially simultaneously or the blocks can sometimes be executed in the reverse order. In addition, the embodiments presented and described in the flowcharts of the present invention are provided by way of example for the purpose of providing a more comprehensive understanding of the technology. The disclosed methods are not limited to the operations and logical flows presented herein. Alternative embodiments are contemplated, where the order of various operations is changed and the sub-operations described as part of a larger operation are executed independently.

[0104] In addition, although the present invention is described in the context of functional modules, it should be understood that unless otherwise stated to the contrary, one or more of the functions and / or features described may be integrated in a single physical device and / or software module, or one or more functions and / or features may be implemented in separate physical devices or software modules. It can also be understood that a detailed discussion of the actual implementation of each module is not necessary for understanding the present invention. Rather, considering the attributes, functions and internal relationships of the various functional modules in the devices disclosed herein, the actual implementation of the module will be understood within the ordinary skills of an engineer. Therefore, those skilled in the art can implement the present invention as set forth in the claims without undue experimentation. It can also be understood that the specific concepts disclosed are merely illustrative and are not intended to limit the scope of the present invention, which is determined by the full scope of the appended claims and their equivalents.

[0105] If the above-mentioned functions are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or a part of this technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in various embodiments of the present invention. The aforementioned storage medium includes: various media that can store program codes, such as USB flash drives, mobile hard disks, read-only memories (ROM, Read-Only Memory), random access memories (RAM, Random Access Memory), magnetic disks, or optical discs.

[0106] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as a definite sequence list of executable instructions for implementing logical functions, and can be specifically implemented in any computer-readable medium for use by an instruction execution system, apparatus, or device (such as a computer-based system, a system including a processor, or other systems that can fetch and execute instructions from the instruction execution system, apparatus, or device), or in conjunction with these instruction execution systems, apparatuses, or devices. For the purposes of this specification, a "computer-readable medium" can be any device that can contain, store, communicate, propagate, or transport a program for use by or in conjunction with an instruction execution system, apparatus, or device.

[0107] More specific examples (non-exhaustive list) of computer-readable media include the following: an electrical connection part with one or more wirings (electronic device), a portable computer disk cartridge (magnetic device), random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), fiber optic device, and portable compact disc read-only memory (CDROM). Additionally, the computer-readable medium can even be paper or other suitable media on which the program can be printed, because the program can be obtained electronically, for example, by optically scanning the paper or other media, then editing, interpreting, or otherwise processing it as necessary, and then storing it in a computer memory.

[0108] It should be understood that various parts of the present invention can be implemented by hardware, software, firmware or a combination thereof. In the above embodiments, multiple steps or methods can be implemented by software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented by hardware, as in another embodiment, any one or a combination of the following techniques well known in the art can be used: discrete logic circuits having logic gate circuits for implementing logic functions on data signals, application specific integrated circuits having appropriate combinational logic gate circuits, programmable gate arrays (PGAs), field programmable gate arrays (FPGAs), etc.

[0109] In the description of this specification, the description with reference to the terms "one embodiment", "some embodiments", "example", "specific example", or "some examples", etc. means that the specific features, structures, materials, or characteristics described in connection with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic representations of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials, or characteristics described can be combined in any one or more embodiments or examples in a suitable manner.

[0110] Although the embodiments of the present invention have been shown and described, those of ordinary skill in the art can understand that various changes, modifications, substitutions, and variations can be made to these embodiments without departing from the principles and spirit of the present invention, and the scope of the present invention is defined by the claims and their equivalents.

[0111] The above is a specific description of the preferred embodiments of the present invention, but the present invention is not limited to the described embodiments. Those skilled in the art can also make various equivalent deformations or substitutions without departing from the spirit of the present invention, and these equivalent deformations or substitutions are all included within the scope defined by the claims of this application.

Claims

1. A noise recognition method, characterized in that, Including: Collecting audio samples respectively from areas only affected by road traffic noise and areas only affected by social life noise to obtain sample audio data; Performing signal processing and feature extraction on the sample audio data to obtain a dataset for training, where the signal processing includes evenly splitting the sample audio data to obtain multiple audio segments of the same time length; Training a deep convolutional neural network model based on the dataset for training to obtain a sound source recognition model; Identifying the noise in the area to be identified according to the sound source recognition model to determine the type of the noise source in the area to be identified; The step of identifying the noise in the area to be identified according to the sound source recognition model to determine the type of the noise source in the area to be identified includes the following steps: Collecting audio data of environmental noise in the area to be identified, filtering the audio data, and generating a Mel logarithmic spectrogram and an MFCC feature map; where the signal processing includes evenly splitting the sample audio data to obtain multiple audio segments of the same time length Inputting the Mel logarithmic spectrogram and the MFCC feature map into the trained sound source recognition model, identifying the features of the spectrogram and the MFCC feature map, and then discriminating the sound source category for the features identified in each time segment to obtain the main sound source category discrimination results for each time segment; Judging the main environmental noise source in the area to be identified according to the comparison of the recognition quantities of road traffic noise and social life noise in the main sound source category discrimination results of different time segments.

2. The noise recognition method according to claim 1, characterized in that The step of collecting audio samples respectively from areas only affected by road traffic noise and areas only affected by social life noise to obtain sample audio data includes: Selecting areas only affected by road traffic noise and areas only affected by social life noise, collecting environmental noise and performing audio recording to obtain sample audio signals; Performing data augmentation operations on the sample audio signals to obtain sample audio data; where the data augmentation operations include operations of adding Gaussian white noise, waveform displacement, and waveform stretching.

3. A noise recognition method according to claim 1, characterized in that, The step of performing signal processing and feature extraction on the sample audio data to obtain a dataset for training includes: Splitting the sample audio data to obtain multiple audio segments with a time length of 5 seconds; Performing short-time Fourier transform on the split audio segments to obtain the spectra of each audio segment; Constructing a Mel filter, filtering the spectra of the audio segments to obtain a Mel logarithmic spectrum; After performing discrete cosine transform on the Mel logarithmic spectrum, selecting Mel cepstral coefficients to obtain MFCC features; Rendering the Mel logarithmic spectrum and the MFCC features into graphs to obtain the Mel logarithmic spectrogram and the MFCC feature map of road traffic noise and social life noise.

4. A noise recognition method according to claim 3, characterized in that, The step of training a deep convolutional neural network model based on the dataset for training to obtain a sound source recognition model includes: Constructing a deep convolutional neural network model through the stacking of two-dimensional convolutional layers and max pooling layers; Using the Mel logarithmic spectrogram and the MFCC feature map of road traffic noise and social life noise as inputs respectively to train the model.

5. The noise recognition method according to claim 3, wherein Performing signal processing and feature extraction on the sample audio data to obtain a dataset to be trained further includes: Determining the Hertz frequency range for audio signal processing according to the audible threshold range of the human ear; Determining the start and end frequencies of the Mel filter according to the correspondence between Mel frequency and Hertz frequency; Constructing a Mel filter with 100 center frequencies, filtering the audio signal, and converting the audio signal with continuous frequencies into discrete audio signals at 100 Mel center frequencies to obtain a Mel log spectrum; Performing discrete cosine transform on the Mel log spectrum and taking the 2nd to 13th coefficients after the transform as Mel cepstral coefficients; Plotting the spectrogram of the Mel log spectrum and the MFCC coefficient feature map, thereby obtaining the Mel log spectrogram and MFCC feature map representing road traffic noise and social life noise respectively.

6. A noise recognition device, characterized in that, It includes: A first module for collecting audio samples from areas affected only by road traffic noise and areas affected only by social life noise respectively to obtain sample audio data; A second module for performing signal processing and feature extraction on the sample audio data to obtain a dataset to be trained, wherein the signal processing includes uniformly segmenting the sample audio data to obtain multiple audio segments with the same time length; A third module for training a deep convolutional neural network model according to the dataset to be trained to obtain a sound source recognition model; A fourth module for identifying the noise in the area to be identified according to the sound source recognition model to determine the type of noise source in the area to be identified; The identifying the noise in the area to be identified according to the sound source recognition model to determine the type of noise source in the area to be identified includes the following steps: Collecting audio data of environmental noise in the area to be identified, filtering the audio data, and generating a Mel log spectrogram and an MFCC feature map; Inputting the Mel log spectrogram and the MFCC feature map into the trained sound source recognition model, identifying the features of the spectrogram and the MFCC feature map, and then discriminating the sound source category for the features identified in each time segment to obtain the main sound source category discrimination results for each time segment; Judging the main environmental noise source in the area to be identified according to the comparison of the recognition quantities of road traffic noise and social life noise in the main sound source category discrimination results of different time segments.

7. An electronic device, characterized in that, It includes a processor and a memory; The memory is used for storing programs; The processor executes the program to implement the method according to any one of claims 1 to 5.

8. A computer-readable storage medium, characterized in that, The storage medium stores a program, and the program is executed by the processor to implement the method according to any one of claims 1 to 5.

9. A computer program product, comprising a computer program, characterized in that, The computer program, when executed by the processor, implements the method according to any one of claims 1 to 5.

Images