Sound classification method and device based on improved African vulture, equipment and storage medium

Abstract

The invention discloses a sound classification method, device and equipment based on improved African vulture, and a storage medium, and relates to the technical field of sound classification, and the method comprises the steps: extracting the Mel frequency cepstral coefficient features of sound data; optimizing the African vulture optimization algorithm according to a preset optimization strategy to obtain a target African vulture optimization algorithm, wherein the preset optimization strategy comprises a Cauchy Gaussian mixture variation strategy, an adaptive weight coefficient strategy and a simplex method strategy; optimizing hyper-parameters of a long and short term memory network model according to the target African vulture optimization algorithm to obtain a target hyper-parameter combination; and reconstructing the long and short-term memory model according to the target hyper-parameter combination and the Mel-frequency cepstral coefficient features to obtain a sound classification model so as to realize sound classification. According to the method, the new sound classification model is constructed by improving the African vulture optimization algorithm, so that the accuracy and the stability of urban sound classification are improved.

Description

Technical Field

[0001] This application relates to the technical field of sound classification, and in particular, to a sound classification method, device, equipment and storage medium based on an improved African vulture. Background Art

[0002] As an important carrier for information transmission, sound signals contain rich environmental and emotional information. Environmental sound, as one of the most common types of sound, plays a crucial role in the field of sound signal processing. Accurate urban sound classification technology can make correct judgments on randomly occurring sounds, so as to give early warnings or take corresponding actions in a timely manner when certain emergencies occur, which is of great significance for fields such as urban safety, traffic management, and environmental monitoring. Traditionally, sound classification mainly relies on traditional machine learning algorithms such as GMM (Gaussian Mixture Model), SVM (Support Vector Machine), and KNN (K-Nearest Neighbor). However, these algorithms have limited generalization ability when faced with the complex and variable sound characteristics in the urban environment. There are a wide variety of sounds in the urban environment, and they have spatio-temporal dynamics. Traditional models often have difficulty effectively capturing these characteristics, resulting in low classification accuracy.

[0003] In recent years, with the rapid development of deep learning technology, especially the remarkable advantages shown by LSTM (Long Short-Term Memory) in processing time-series data, it provides a new solution idea for sound classification. LSTM is a special variant of RNN (Recurrent Neural Network). By introducing the gating mechanism of input gate, forget gate and output gate, it effectively solves the deficiency of traditional RNN in capturing long-term dependence relationships. In the sound classification task, LSTM can make full use of the time-series characteristics of sound signals, extract deeper features, and thus improve the classification accuracy. However, the performance of LSTM highly depends on its parameter settings, including hyperparameters such as the number of network layers, the number of neurons, the learning rate, and the dropout rate. The selection of these parameters directly affects the training effect and generalization ability of the model. Therefore, optimizing the parameters of the LSTM model is the key to improving the sound classification accuracy. AVOA (African Vulture Optimization Algorithm) is a new type of metaheuristic optimization algorithm proposed by Benyamin et al. in 2021. This algorithm is known for its high precision and effectiveness and has been successfully applied to many fields such as image segmentation, portfolio optimization, and parameter optimization. The AVOA algorithm realizes the iterative update of the population and global search by simulating the foraging behavior of African vultures. However, the traditional AVOA algorithm has deficiencies such as poor population diversity, weak global search ability, and being easily trapped in local optimal solutions, which limits its application in complex optimization problems.

[0004] Therefore, how to overcome the deficiencies of traditional sound classification algorithms and the parameter optimization of the LSTM model, and improve the accuracy and stability of urban sound classification, is an urgent problem to be solved at present. Summary of the Invention

[0005] The main purpose of this application is to provide a sound classification method, device, equipment and storage medium based on improved African vultures, aiming to solve the technical problem of how to overcome the deficiencies of traditional sound classification algorithms and the parameter optimization of the LSTM model, and improve the accuracy and stability of urban sound classification.

[0006] To achieve the above purpose, this application proposes a sound classification method based on improved African vultures, and the method includes:

[0007] Extract the Mel Frequency Cepstral Coefficient (MFCC) features of the sound data;

[0008] Optimize the African Vulture Optimization Algorithm according to a preset optimization strategy to obtain a target African Vulture Optimization Algorithm, and the preset optimization strategy includes a Cauchy-Gaussian hybrid mutation strategy, an adaptive weight coefficient strategy, and a simplex method strategy;

[0009] Optimize the hyperparameters of the long short - term memory network model according to the target African vulture optimization algorithm to obtain a target hyperparameter combination;

[0010] Reconstruct the long short - term memory model according to the target hyperparameter combination and the mel - frequency cepstral coefficient features to obtain a sound classification model for realizing sound classification.

[0011] In one embodiment, the step of optimizing the African vulture optimization algorithm according to a preset optimization strategy to obtain a target African vulture optimization algorithm includes:

[0012] Obtain the starvation rate of the current African vulture optimization algorithm;

[0013] Determine the algorithm operation stage according to the range of the starvation rate;

[0014] Optimize the African vulture optimization algorithm according to the Cauchy - Gaussian mixed mutation strategy, the adaptive weight coefficient strategy, and the simplex method strategy according to the algorithm operation stage to obtain a target African vulture optimization algorithm.

[0015] In one embodiment, the step of optimizing the African vulture optimization algorithm according to the Cauchy - Gaussian mixed mutation strategy, the adaptive weight coefficient strategy, and the simplex method strategy according to the algorithm operation stage to obtain a target African vulture optimization algorithm includes:

[0016] When the algorithm operation stage is the global exploration stage, optimize the African vulture optimization algorithm according to the Cauchy - Gaussian mixed mutation strategy to obtain a first optimization equation;

[0017] When the algorithm operation stage is the local development stage, optimize the African vulture optimization algorithm according to the adaptive weight coefficient strategy to obtain a second optimization equation;

[0018] When the algorithm operation stage is the global optimization stage, optimize the African vulture optimization algorithm according to the simplex method strategy to obtain a third optimization equation;

[0019] Based on the African vulture optimization algorithm, the first optimization equation, the second optimization equation, and the third optimization equation, obtain a target African vulture optimization algorithm.

[0020] In one embodiment, the step of optimizing the African vulture optimization algorithm according to the Cauchy - Gaussian mixed mutation strategy to obtain a first optimization equation when the algorithm operation stage is the global exploration stage includes:

[0021] When the algorithm operation stage is the global exploration stage, determine the adaptive weight and the greedy mechanism;

[0022] The first optimization equation is obtained by adding the adaptive weight and the greedy mechanism to the African vulture optimization algorithm according to the Cauchy-Gaussian mixture mutation strategy.

[0023] In one embodiment, when the algorithm running stage is the local development stage, the steps of optimizing the African vulture optimization algorithm according to the adaptive weight coefficient strategy to obtain the second optimization equation include:

[0024] When the algorithm running stage is the local development stage, introduce the adaptive weight coefficient according to the adaptive weight coefficient strategy;

[0025] Add the adaptive weight coefficient to the African vulture optimization algorithm to obtain the second optimization equation.

[0026] In one embodiment, when the algorithm running stage is the global optimization stage, the steps of optimizing the African vulture optimization algorithm according to the simplex method strategy to obtain the third optimization equation include:

[0027] When the algorithm running stage is the global optimization stage, sort the population individuals of the African vulture optimization algorithm to obtain the first vulture position and the second vulture position;

[0028] Obtain the midpoint position according to the first vulture position and the second vulture position;

[0029] Perform reflection, expansion, outer contraction or inner contraction operations based on the simplex method strategy and the midpoint position to obtain the third optimization equation.

[0030] In one embodiment, the steps of optimizing the hyperparameters of the long short-term memory network model according to the target African vulture optimization algorithm to obtain the target hyperparameter combination include:

[0031] Set the hyperparameters of the long short-term memory network model, and the hyperparameters include the number of neurons in the first layer, the number of neurons in the second layer, the learning rate, and the dropout rate;

[0032] Map the hyperparameters to the target African vulture optimization algorithm, and iteratively optimize through the target African vulture optimization algorithm to obtain the target hyperparameter combination.

[0033] In addition, to achieve the above object, the present application also proposes a voice classification device based on the improved African vulture, and the device includes:

[0034] A feature extraction module, configured to extract the Mel frequency cepstral coefficient features of the voice data;

[0035] An algorithm optimization module, configured to optimize the African vulture optimization algorithm according to a preset optimization strategy to obtain a target African vulture optimization algorithm, where the preset optimization strategy includes a Cauchy-Gaussian hybrid mutation strategy, an adaptive weight coefficient strategy, and a simplex method strategy;

[0036] A model optimization module, configured to optimize the hyperparameters of a long short-term memory network model according to the target African vulture optimization algorithm to obtain a target hyperparameter combination;

[0037] A sound classification module, configured to reconstruct the long short-term memory model according to the target hyperparameter combination and the Mel frequency cepstral coefficient features to obtain a sound classification model for realizing sound classification.

[0038] In addition, to achieve the above object, the present application also provides a sound classification device based on an improved African vulture. The device includes: a memory, a processor, and a computer program stored on the memory and executable on the processor. The computer program is configured to implement the steps of the sound classification method based on the improved African vulture as described above.

[0039] In addition, to achieve the above object, the present application also provides a storage medium. The storage medium is a computer-readable storage medium, and a computer program is stored on the storage medium. When the computer program is executed by a processor, the steps of the sound classification method based on the improved African vulture as described above are implemented.

[0040] In addition, to achieve the above object, the present application also provides a computer program product. The computer program product includes a computer program. When the computer program is executed by a processor, the steps of the sound classification method based on the improved African vulture as described above are implemented.

[0041] The present application provides a sound classification method based on an improved African vulture. The method of the present application includes: extracting Mel frequency cepstral coefficient features of sound data; optimizing the African vulture optimization algorithm according to a preset optimization strategy to obtain a target African vulture optimization algorithm, where the preset optimization strategy includes a Cauchy-Gaussian hybrid mutation strategy, an adaptive weight coefficient strategy, and a simplex method strategy; optimizing the hyperparameters of a long short-term memory network model according to the target African vulture optimization algorithm to obtain a target hyperparameter combination; reconstructing the long short-term memory model according to the target hyperparameter combination and the Mel frequency cepstral coefficient features to obtain a sound classification model for realizing sound classification. In summary, it can be seen that the present application improves the accuracy and stability of urban sound classification by improving the African vulture optimization algorithm and constructing a new sound classification model. Description of the Drawings

[0042] The accompanying drawings here are incorporated into the specification and form a part of this specification, showing embodiments consistent with this application, and are used together with the specification to explain the principles of this application.

[0043] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the following will briefly introduce the accompanying drawings required for use in the description of the embodiments or the prior art. Obviously, for those of ordinary skill in the art, without creative efforts, other drawings can also be obtained based on these drawings.

[0044] Figure 1 It is a schematic flowchart provided for the first embodiment of the method for classifying the sounds of African vultures based on improvements in this application;

[0045] Figure 2 It is a Mel-frequency cepstral coefficient feature map in an embodiment of the method for classifying the sounds of African vultures based on improvements in this application;

[0046] Figure 3 It is a schematic flowchart of the Mel-frequency cepstral coefficient feature extraction in an embodiment of the method for classifying the sounds of African vultures based on improvements in this application;

[0047] Figure 4 It is a schematic flowchart of the sound classification model in an embodiment of the method for classifying the sounds of African vultures based on improvements in this application;

[0048] Figure 5 It is a schematic flowchart provided for the second embodiment of the method for classifying the sounds of African vultures based on improvements in this application;

[0049] Figure 6 It is a graph of the change in the starvation rate in an embodiment of the method for classifying the sounds of African vultures based on improvements in this application;

[0050] Figure 7 It is a graph of the standard Gaussian distribution and the standard Cauchy distribution in an embodiment of the method for classifying the sounds of African vultures based on improvements in this application;

[0051] Figure 8 It is the w in an embodiment of the method for classifying the sounds of African vultures based on improvements in this application 1 Graph of the change trend of the weight coefficient;

[0052] Figure 9 It is the w in an embodiment of the method for classifying the sounds of African vultures based on improvements in this application 2 Graph of the change trend of the weight coefficient;

[0053] Figure 10 It is a schematic diagram of the module structure of the device for classifying the sounds of African vultures based on improvements in the embodiments of this application;

[0054] Figure 11It is a schematic diagram of the device structure of the hardware operating environment involved in the voice classification method based on the improved African vulture in the embodiments of the present application.

[0055] The implementation, functional characteristics and advantages of the present application will be further described with reference to the accompanying drawings in combination with the embodiments. Specific implementation manners

[0056] It should be understood that the specific embodiments described herein are only used to explain the technical solutions of the present application and are not used to limit the present application.

[0057] To better understand the technical solutions of the present application, the following will be described in detail in combination with the accompanying drawings of the specification and specific implementation manners.

[0058] The main solution of the embodiments of the present application is: extracting the Mel-frequency cepstral coefficient features of the voice data; optimizing the African vulture optimization algorithm according to a preset optimization strategy to obtain a target African vulture optimization algorithm, where the preset optimization strategy includes a Cauchy-Gaussian mixture mutation strategy, an adaptive weight coefficient strategy and a simplex method strategy; optimizing the hyperparameters of the long short-term memory network model according to the target African vulture optimization algorithm to obtain a target hyperparameter combination; reconstructing the long short-term memory model according to the target hyperparameter combination and the Mel-frequency cepstral coefficient features to obtain a voice classification model to implement voice classification.

[0059] As an important carrier for information transmission, voice signals contain rich environmental and emotional information. Environmental sounds, as one of the most common types of sounds, play a crucial role in the field of voice signal processing. Accurate urban voice classification technology can make correct judgments on randomly occurring sounds, so as to give early warnings or take corresponding actions in a timely manner when certain emergencies occur, which is of great significance for fields such as urban security, traffic management, and environmental monitoring. Traditionally, voice classification mainly relied on traditional machine learning algorithms such as GMM, SVM, and KNN. However, these algorithms have limited generalization ability when faced with the complex and changeable voice features in the urban environment. There are a wide variety of sounds in the urban environment and they have spatio-temporal dynamics. Traditional models often have difficulty effectively capturing these features, resulting in low classification accuracy.

[0060] In recent years, with the rapid development of deep learning technology, especially the remarkable advantages shown by LSTM in processing time series data, it has provided new solutions for sound classification. LSTM is a special variant of RNN. By introducing the gating mechanism of input gate, forget gate and output gate, it effectively solves the deficiency of traditional RNN in capturing long-term dependence relationships. In the sound classification task, LSTM can make full use of the time series characteristics of sound signals, extract deeper features, and thus improve the classification accuracy. However, the performance of LSTM highly depends on its parameter settings, including hyperparameters such as the number of network layers, the number of neurons, the learning rate, and the dropout rate. The selection of these parameters directly affects the training effect and generalization ability of the model. Therefore, optimizing the parameters of the LSTM model is the key to improving the sound classification accuracy. AVOA is a new type of meta-heuristic optimization algorithm proposed by Benyamin et al. in 2021. This algorithm is known for its high accuracy and effectiveness and has been successfully applied in many fields such as image segmentation, portfolio optimization, and parameter optimization. The AVOA algorithm realizes the iterative update of the population and global search by simulating the foraging behavior of African vultures. However, the traditional AVOA algorithm has deficiencies such as poor population diversity, weak global search ability, and being prone to falling into local optimal solutions, which limits its application in complex optimization problems. Therefore, how to improve the accuracy and stability of urban sound classification is an urgent problem to be solved currently.

[0061] This application overcomes the deficiencies of traditional sound classification algorithms and LSTM model parameter optimization by improving the African vulture optimization algorithm and constructing a new sound classification model, and improves the accuracy and stability of urban sound classification.

[0062] It should be noted that the execution subject of this embodiment can be a sound classification system based on the improved African vulture, or a computing service device with data processing, network communication, and program running functions, such as a tablet computer, a personal computer, a mobile phone, etc., or an electronic device capable of implementing the above-mentioned sound classification function based on the improved African vulture. This embodiment does not specifically limit this. Hereinafter, taking the sound classification system based on the improved African vulture as an example, this embodiment and the following embodiments will be described.

[0063] Based on this, the embodiment of this application provides a sound classification method based on the improved African vulture, referring to Figure 1 , Figure 1 is the flow chart of the first embodiment of the sound classification method based on the improved African vulture of this application.

[0064] In this embodiment, the sound classification method based on the improved African vulture includes steps S10 to S40:

[0065] Step S10: Extract the Mel-frequency cepstral coefficient features of the sound data.

[0066] It should be noted that, as Figure 2 shown, the Mel Frequency Cepstral Coefficient (MFCC) feature is a feature that can reflect the energy distribution of the sound signal in the Mel frequency domain and the auditory perception characteristics of the human ear, and has significant discrimination ability for the sound classification task. Specifically, as Figure 3 shown, the system will first preprocess the original speech signal, including operations such as pre-emphasis, framing and windowing, and fast discrete Fourier transform, to obtain a series of short-time sound signals; then, define a filter bank to convert each short-time sound signal into a representation in the Mel frequency domain, usually implemented through a Mel filter bank; then, perform a logarithmic transformation on the output of the Mel filter bank and perform a discrete cosine transform (i.e., DCT) on it to obtain the MFCC feature vector. For example, in the scenario of urban sound classification, the UrbanSound8K dataset can be used as experimental data. This dataset contains various urban sound samples, such as car sounds, children playing sounds, dog barking sounds, etc. For each sound sample, extract its MFCC feature according to the above process to construct a feature vector sequence, providing input data for the subsequent classification task.

[0067] Step S20: Optimize the African Vulture Optimization Algorithm according to a preset optimization strategy to obtain the target African Vulture Optimization Algorithm, where the preset optimization strategy includes a Cauchy-Gaussian hybrid mutation strategy, an adaptive weight coefficient strategy, and a simplex method strategy.

[0068] It should be noted that this step mainly aims at the problems of poor population diversity, weak global search ability, and easy to fall into local optimal solutions existing in the African Vulture Optimization Algorithm (AVOA), and proposes a preset optimization strategy that combines the Cauchy-Gaussian hybrid mutation strategy, the adaptive weight coefficient strategy, and the simplex method strategy to effectively improve the AVOA algorithm.

[0069] In addition, it should be noted that in the AVOA algorithm, when |F|≥1 (i.e., the starvation rate), the algorithm enters the global exploration stage, when |When F < 1, the algorithm enters the local development stage. In this embodiment, the Cauchy-Gaussian hybrid mutation strategy refers to introducing the Cauchy-Gaussian hybrid mutation strategy during the global exploration stage of the AVOA algorithm to enhance the global search ability of the algorithm. Specifically, for each individual in the population, a Cauchy mutation term and a Gaussian mutation term are generated based on its current position, and then these two mutation terms are linearly combined with the individual position to generate a new individual position. By introducing Cauchy mutation and Gaussian mutation, a larger search step size can be obtained at the initial stage of algorithm iteration to avoid falling into local optimal solutions; while in the later stage of iteration, by adjusting the weight of Gaussian mutation, local perturbation of the current position can be realized to improve the convergence accuracy of the algorithm. The adaptive weight coefficient strategy refers to designing an adaptive weight coefficient during the local development stage of the AVOA algorithm to strengthen the guiding role of the optimal vulture and sub-optimal vulture individuals. Specifically, according to the current iteration number and individual fitness value, the weight coefficients of the optimal vulture and sub-optimal vulture are dynamically adjusted to update the individual position. At the initial stage of iteration, the guiding role of the sub-optimal vulture is enhanced to improve the global search ability; while in the later stage of iteration, the guiding role of the optimal vulture is gradually enhanced to prompt the population to converge to the optimal solution as soon as possible. The simplex method strategy refers to introducing the simplex method to optimize the positions of individuals with poor fitness in order to further improve the search performance of the algorithm. Specifically, for individuals with poor fitness in the population, the simplex method is used for local search to find better individual positions. By continuously iterating and updating the individual positions, the global optimal solution can be gradually approximated. Through comparative experiments, it can be verified that the target AVOA algorithm (i.e., MY_AVOA algorithm) integrating the above three strategies has significant improvements in convergence speed, ability to jump out of local optima, and convergence accuracy.

[0070] Step S30: Optimize the hyperparameters of the long short-term memory network model according to the target African vulture optimization algorithm to obtain a target hyperparameter combination.

[0071] It should be noted that in this step, the system will use the target African vulture optimization algorithm (MY_AVOA) to optimize the hyperparameters of the long short-term memory network (LSTM) model to obtain the optimal target hyperparameter combination. It can be understood that the LSTM model is a variant of the recurrent neural network for processing sequential data, and its performance highly depends on the setting of hyperparameters, such as the number of neurons, learning rate, dropout rate, etc.

[0072] In a feasible implementation manner, the step S30 specifically includes:

[0073] Step S301: Set the hyperparameters of the long short-term memory network model, and the hyperparameters include the number of neurons in the first layer, the number of neurons in the second layer, the learning rate, and the dropout rate.

[0074] It should be noted that in this step, the hyperparameters of the long short-term memory (LSTM) model are the key to optimizing the model performance, which can directly affect the training efficiency and classification accuracy of the model. Specifically, the hyperparameters include the number of neurons in the first layer (neurons1) and the number of neurons in the second layer (neurons2). The number of neurons determines the learning ability and complexity of the model. Too many neurons may lead to overfitting, while too few neurons may not be able to fully learn the features of the data. Secondly, the learning rate (lr). The learning rate controls the step size of parameter updates during the training process. A too large learning rate may cause the model to oscillate near the optimal solution, while a too small learning rate will make the training process slow. The dropout rate (dropoutrate), the dropout rate is a regularization technique used to prevent model overfitting. During the training process, randomly discard the outputs of some neurons with a certain probability, which can increase the generalization ability of the model.

[0075] Step S302: Map the hyperparameters to the target African vulture optimization algorithm, and iteratively optimize through the target African vulture optimization algorithm to obtain a target hyperparameter combination.

[0076] It should be noted that as Figure 3 shown, in this step, the system will map the hyperparameters of the LSTM model to the target African vulture optimization algorithm (MY_AVOA), and perform iterative optimization to obtain the optimal hyperparameter combination. Specifically, take the number of neurons in the first layer, the number of neurons in the second layer, the learning rate, and the dropout rate of the LSTM model as the positions of the African vultures, and take the classification accuracy of the LSTM model as the fitness function. In the MY_AVOA algorithm, each vulture represents a potential solution, that is, a set of hyperparameter combinations. Then, by running the MY_AVOA algorithm, iteratively optimize the global optimal position of the vultures, and continuously adjust the hyperparameter combination of the LSTM model. In each iteration, update the position of the vultures according to the update rule of the MY_AVOA algorithm, that is, update the hyperparameter combination of the LSTM model, and calculate the fitness function value (i.e., the classification accuracy of the LSTM model). Through continuous iteration, the algorithm gradually approaches the optimal solution. Finally, when the maximum number of iterations is reached, output the optimal vulture position, that is, the target hyperparameter combination.

[0077] Step S40: Reconstruct the long short-term memory model according to the target hyperparameter combination and the mel-frequency cepstral coefficient features to obtain a sound classification model to achieve sound classification.

[0078] It should be noted that as Figure 4As shown, in this step, the system takes the extracted Mel Frequency Cepstral Coefficient features (i.e., MFCC features) as input data and inputs them into a Long Short-Term Memory model (LSTM model) to reconstruct the LSTM model. By training the reconstructed LSTM model, it can learn the mapping relationship between sound features and sound categories. Finally, the trained LSTM model is used to classify the sound data in the test set, and the classification performance is evaluated. For example, according to the target hyperparameter combination, set the number of neurons in the first layer, the number of neurons in the second layer, the learning rate, and the dropout rate of the LSTM model. Then, input the extracted MFCC features into the LSTM model in the order of time series. Through multiple iterations of training, the LSTM model can accurately identify different categories of sounds. Finally, use the trained MY_AVOA-LSTM model (sound classification model) to classify the sound data in the test set. And the performance of the model can be evaluated by calculating classification metrics such as accuracy, precision, recall, and F1-score.

[0079] It can be understood that in order to verify the superiority of the MY_AVOA-LSTM model for urban sound classification problems, five algorithms, namely SSA, WOA, PSO, AVOA, and MY_AVOA, are simultaneously selected to optimize the parameters of the LSTM. The population size is set to 10, the maximum number of iterations is 30, the range of the number of neurons neurons1 and neurons2 in the LSTM is [5, 512], the range of the learning rate lr is [0.001, 0.1], and the range of the dropout rate is [0.2, 0.5]. The optimization results of each algorithm for the LSTM parameters are shown in Table 1.

[0080] Table 1

[0081]

[0082] Based on the LSTM parameter combinations obtained through optimization, SSA-LSTM, WOA-LSTM, PSO-LSTM, AVOA-LSTM, and MY_AVOA-LSTM urban sound classification models are respectively established, and LightGBM, MLP, XGBoost, and CNN models are introduced for horizontal comparison and verification. The results of the classification accuracies of different models are shown in Table 2.

[0083] Table 2

[0084] Classification model Accuracy Precision Recall F1-score LightGBM 0.789 0.811 0.780 0.790 MLP 0.713 0.756 0.711 0.725 XGBoost 0.774 0.794 0.764 0.774 CNN 0.781 0.803 0.795 0.784 LSTM 0.813 0.811 0.827 0.815 SSA-LSTM 0.825 0.838 0.837 0.835 WOA-LSTM 0.871 0.876 0.878 0.875 PSO-LSTM 0.816 0.831 0.821 0.823 AVOA-LSTM 0.915 0.913 0.919 0.916 MY_AVOA-LSTM 0.937 0.940 0.941 0.940

[0085] As can be seen from Table 2, the accuracy of the MY_AVOA-LSTM model can reach 93.7%. Compared with the SSA-LSTM, WOA-LSTM, PSO-LSTM, and AVOA-LSTM models, the accuracy of MY_AVOA-LSTM has increased by 11.2%, 6.6%, 12.1%, and 2.2% respectively, indicating that the improvement of the MY_AVOA algorithm is effective. Compared with other models, the MY_AVOA-LSTM model shows a lower classification error rate, improving the accuracy and stability of the sound classification model.

[0086] This embodiment provides a sound classification method based on an improved African vulture optimization algorithm. The method of this embodiment includes: extracting the Mel-frequency cepstral coefficient features of sound data; optimizing the African vulture optimization algorithm according to a preset optimization strategy to obtain a target African vulture optimization algorithm, where the preset optimization strategy includes a Cauchy-Gaussian mixed mutation strategy, an adaptive weight coefficient strategy, and a simplex method strategy; optimizing the hyperparameters of the long short-term memory network model according to the target African vulture optimization algorithm to obtain a target hyperparameter combination; and reconstructing the long short-term memory model according to the target hyperparameter combination and the Mel-frequency cepstral coefficient features to obtain a sound classification model to achieve sound classification. In summary, this application improves the accuracy and stability of urban sound classification by improving the African vulture optimization algorithm and constructing a new sound classification model.

[0087] Based on the first embodiment of this application, in the second embodiment of this application, the same or similar content as in the above-mentioned first embodiment can be referred to the above introduction and will not be repeated hereinafter. On this basis, please refer to Figure 5 , Figure 5 which is a schematic flowchart of the second embodiment of the sound classification method based on the improved African vulture of this application. The specific steps of S20 include:

[0088] Step S201: Obtain the starvation rate of the current African vulture optimization algorithm.

[0089] It should be noted that the starvation rate F of the African vulture optimization algorithm is used to control the conversion between the exploration stage and the exploitation stage of the algorithm. Its calculation formula is shown in Formula 1:

[0090]

[0091] where It represents the current iteration number, T max represents the maximum iteration number, z is a random number that iteratively changes within the interval [-1, 1], h is a random number within the interval [-2, 2], rand 1 is a random number within the interval (0, 1), and w is the exploitation probability with a value of 2.5. Figure 6 is the change situation of the starvation rate F.

[0092] Step S202: Determine the algorithm running stage according to the range of the starvation rate.

[0093] It should be noted that in this embodiment, according to the range of the starvation rate F, the running stage of the algorithm can be divided into a global exploration stage and a local exploitation stage. When |F|≥1, the algorithm enters the global exploration stage. At this time, the individual position update method is shown in Equation 2:

[0094]

[0095] where P(i + 1) is the position of the vulture in the next iteration, P 1 is a preset exploration parameter used to control the exploration strategy, P 1 = 0.6, rand p1 , rand 2 and rand 3 are all random numbers within the interval (0, 1), and ub and lb are the upper and lower bounds of the search range.

[0096] When |F| < 1, the algorithm enters the local exploitation stage. This stage is further divided into the following two stages:

[0097] When 0.5 < |F| < 1, it is determined by P 2 to execute the siege strategy or the spiral flight strategy, and the position update method is shown in Equation 3:

[0098]

[0099] where P 2 is a preset selection parameter with a value of 0.4, rand p2 , rand 4 , rand 5 and rand 6 are all random numbers between (0, 1).

[0100] When 0 < |F| < 0.5, it is determined by P 3 to execute the siege strategy or the aggregation strategy, and the position update method is as shown in Equation 4:

[0101] as shown in Equation 4:

[0102]

[0103] where P 3 is a preset selection parameter, P 3 = 0.6, BestVulture 1 (i) is the first group of best vultures in the current iteration, BestVulture 2(i) is the second group of the best vultures in the current iteration, P(i) is the position vector of the current vulture, and d(t) represents the distance between the vulture and the best vulture in the two groups.

[0104] It can be understood that by determining the running stage of the algorithm according to the range of the hunger rate F, the algorithm can adaptively adjust its search strategy, thereby achieving a balance between global exploration and local exploitation and improving the search efficiency and the quality of the solution.

[0105] In addition, it should be noted that the running stage of the algorithm also includes a global optimization stage. The global optimization stage refers to the later stage of the African Vulture Optimization Algorithm (AVOA algorithm). The role of the simplex method is to optimize the African vulture individuals with poor fitness during the algorithm iteration to avoid the algorithm falling into a stagnant state and improve the search performance.

[0106] Step S203: Optimize the African Vulture Optimization Algorithm according to the running stage of the algorithm by invoking the Cauchy-Gaussian hybrid mutation strategy, the adaptive weight coefficient strategy, and the simplex method strategy to obtain the target African Vulture Optimization Algorithm.

[0107] It should be noted that in this step, the system will optimize the African Vulture Optimization Algorithm by invoking different Cauchy-Gaussian hybrid mutation strategies, adaptive weight coefficient strategies, and simplex method strategies according to the running stage of the algorithm. Specifically, in the global exploration stage, the Cauchy-Gaussian hybrid mutation strategy is mainly used to enhance the global search ability of the algorithm. By introducing the combination of Cauchy mutation and Gaussian mutation, the algorithm can generate larger step sizes during the search process, thus making it easier to jump out of the local optimal solution and discover new search regions. In the local exploitation stage, the adaptive weight coefficient strategy is mainly used. The adaptive weight coefficient strategy makes the algorithm converge to the optimal solution faster in the later stage of iteration by dynamically adjusting the guiding roles of the optimal vulture and the sub-optimal vulture. In the global optimization stage, the simplex method strategy is mainly used. The simplex method further improves the search performance and convergence speed of the algorithm by performing operations such as reflection, expansion, outer contraction, and inner contraction on the vulture individuals with poor fitness.

[0108] It can be understood that by invoking different optimization strategies in different stages, the algorithm can search the solution space more efficiently, improve the search efficiency and the quality of the solution. At the same time, the introduction of these optimization strategies also makes the algorithm have stronger adaptability and robustness.

[0109] In a feasible implementation manner, the step S203 specifically includes:

[0110] Step A10: When the running stage of the algorithm is the global exploration stage, optimize the African Vulture Optimization Algorithm according to the Cauchy-Gaussian hybrid mutation strategy to obtain the first optimization equation.

[0111] It should be noted that during the exploration stage of AVOA, each vulture will randomly search in the environment to find areas that can satisfy its hunger. However, if the current optimal position is not the theoretical optimal position, the population may fall into a local optimum and stagnate. To enhance the algorithm's comprehensive exploration ability of the solution space, a Cauchy-Gaussian hybrid mutation strategy is introduced to expand the search range and increase the possibility of finding the global optimal solution. It can be understood that the Gaussian distribution is a common continuous probability distribution, and its probability density function is shown in Formula 5:

[0112]

[0113] where σ represents the variance and μ represents the mean. The standard Gaussian distribution follows σ = 0 and μ = 1.

[0114] The Cauchy distribution is a continuous probability distribution with no existing mathematical expectation, and its probability density function is shown in Formula 6:

[0115]

[0116] where x 0 is the location parameter that defines the position of the distribution score, and γ is the scale parameter of the distribution, which controls the width of the distribution. The standard Cauchy distribution follows x 0 = 0 and γ = 1. The graphs of the standard Gaussian distribution (Gaussian) and the standard Cauchy distribution (Cauchy) are as Figure 7 shown. As Figure 7 can be seen, compared with the Gaussian distribution, the Cauchy distribution is more concentrated, and its two sides are more discrete, which indicates that the Cauchy distribution can have a higher probability of generating a random number far from the center point, thus being more likely to jump out of the local optimum and having stronger global exploration ability. While the Gaussian distribution tends to generate random numbers near the center point, which helps to search for better solutions near the local optimum point. It can be understood that in this step, the system will introduce various mechanisms (such as the adaptive weight mechanism and the greedy mechanism) based on the Cauchy-Gaussian hybrid mutation to modify parts of the African vulture optimization algorithm to obtain the first optimization equation.

[0117] In a feasible implementation manner, step A10 specifically includes:

[0118] Step A101: When the algorithm running stage is the global exploration stage, determine the adaptive weight and the greedy mechanism.

[0119] It should be noted that when the algorithm running stage is determined to be the global exploration stage (i.e., when the absolute value of the starvation rate F is greater than or equal to 1), in order to more effectively search for the optimal solution in the vast solution space, the system will introduce an adaptive weight and a greedy mechanism. Specifically, the introduction of the adaptive weight includes: dynamically adjusting the weights of Cauchy mutation and Gaussian mutation by introducing the adaptive weight w, and excluding the case where the weight value is 1. The introduction of the greedy mechanism includes: after each position update, by comparing the individual fitness values before and after mutation, a greedy selection strategy is adopted to retain the better individual position. This mechanism ensures that the algorithm can move towards a better solution space in each iteration, avoiding unnecessary performance degradation.

[0120] Step A102: Add the adaptive weight and the greedy mechanism to the African vulture optimization algorithm according to the Cauchy-Gaussian hybrid mutation strategy to obtain a first optimization equation.

[0121] It should be noted that in this step, the system will perform Cauchy-Gaussian hybrid mutation on the position of each vulture. During the mutation process, the weights of Cauchy mutation and Gaussian mutation are dynamically adjusted according to the adaptive weight w to balance the global search ability and local development ability of the algorithm. Then, the mutated position is substituted into the adaptive weight and the greedy mechanism to obtain the first optimization equation. This equation describes the dynamic process of the vulture position changing with the number of iterations in the global exploration stage, specifically as shown in Formula 7:

[0122]

[0123] Among them, P new (i + 1) is the individual perturbed by the linear Cauchy-Gaussian mutation strategy, and w is the inertia weight. It can be understood that in the initial stage of algorithm iteration, a larger step size can be obtained through a larger Cauchy mutation weight value, effectively avoiding the algorithm falling into a local optimal solution; in the later stage of iteration, increasing the weight value of Gaussian mutation is beneficial to generating a small step size and performing local perturbation on the current position, thereby avoiding premature convergence to a local optimal and improving the convergence accuracy of the algorithm.

[0124] Step A20: When the algorithm running stage is the local development stage, optimize the African vulture optimization algorithm according to the adaptive weight coefficient strategy to obtain a second optimization equation.

[0125] It should be noted that in order to improve the performance of the African vulture optimization algorithm (AVOA) in the local development stage, the system introduces an adaptive weight coefficient strategy. This strategy can dynamically adjust the influence weights of the optimal vulture and the sub-optimal vulture on the position update of ordinary vultures, thereby accelerating the convergence speed of the algorithm towards the optimal solution and enhancing the local development ability of the algorithm.

[0126] In a feasible implementation manner, the specific content of the step A20 includes:

[0127] Step A201: When the algorithm is in the local development stage during the algorithm running phase, introduce the adaptive weight coefficients according to the adaptive weight coefficient strategy.

[0128] It should be noted that in this step, when the algorithm enters the local development stage (i.e., 0 ≤ |F| < 1), two adaptive weight coefficients w 1 and w 2 will be introduced according to the adaptive weight coefficient strategy. These two weight coefficients are respectively used to control the influence degree of the optimal vulture and the sub-optimal vulture on the position update of the ordinary vulture. Specifically, w 1 and w 2 can be dynamically adjusted according to the number of iterations and the preset control factor k.

[0129] Step A202: Add the adaptive weight coefficients to the African vulture optimization algorithm to obtain the second optimization equation.

[0130] It should be noted that the adaptive weight coefficients w 1 and w 2 are added to the African vulture optimization algorithm, thus obtaining the second optimization equation. Specifically, when the algorithm is in the local development stage (i.e., 0 < |F| < 0.5), the positions of the vultures are updated according to the adaptive weight coefficients w 1 and w 2 . At this time, the position update formula (i.e., the second optimization equation) is as shown in Formula 8:

[0131]

[0132] where w 1 , w 2 are weight coefficients, k is the control factor, and k = 8. Figure 8 and Figure 9 are the change trends of the two weight coefficients. That is, in the early stage of iteration, the guiding role of the sub-optimal vulture is enhanced (i.e., the value of w 2 is increased), the global search ability is improved, and the algorithm is prevented from falling into local optimum. In the later stage of iteration, the guiding role of the optimal vulture is gradually enhanced (i.e., the value of w 1 is increased), while the guiding role of the sub-optimal vulture is gradually weakened, so as to prompt the vultures to converge to the optimal solution as soon as possible, thereby improving the algorithm convergence speed and enhancing the algorithm local development ability.

[0133] Step A30: When the algorithm is in the global optimization stage during the algorithm running phase, optimize the African vulture optimization algorithm according to the simplex method strategy to obtain the third optimization equation.

[0134] It should be noted that the simplex method is a direct and fast polyhedron search algorithm, which has the characteristics of fast convergence and wide application and performs excellently in multi-dimensional optimization problems. It can be understood that in order to further enhance the search ability of the African Vulture Optimization Algorithm (AVOA) in the global optimization stage (i.e., the later stage of the AVOA algorithm) and avoid the algorithm falling into a local optimal solution, the simplex method strategy is introduced in the later stage of the African Vulture Optimization Algorithm to optimize it.

[0135] In a feasible implementation manner, the step A30 specifically includes:

[0136] Step A301: When the algorithm running stage is the global optimization stage, sort the population individuals of the African Vulture Optimization Algorithm to obtain the first vulture position and the second vulture position.

[0137] It should be noted that when the algorithm enters the global optimization stage, first sort all individuals in the current population according to their fitness values. The fitness value reflects the quality of each individual in the solution space and is an important indicator for evaluating the performance of individuals. After sorting, the individual with the highest fitness value is marked as the first vulture position (i.e., the optimal position x best ), and the individual with the second highest fitness value is marked as the second vulture position (i.e., the sub-optimal position x next ).

[0138] Step A302: Obtain the midpoint position according to the first vulture position and the second vulture position.

[0139] It should be noted that after obtaining the first vulture position and the second vulture position, it is necessary to calculate the midpoint position between them. The calculation of the midpoint position x m is shown in Formula 9:

[0140] x m =(x best -x next ) / 2

[0141] (Formula 9)

[0142] Step A303: Perform reflection, expansion, outside contraction or inside contraction operations based on the simplex method strategy and the midpoint position to obtain the third optimization equation.

[0143] It should be noted that after obtaining the midpoint position, the algorithm enters the simplex method operation stage. The simplex method is a direct search algorithm that constructs a polyhedron (i.e., simplex) in the solution space and continuously adjusts the shape and position of the polyhedron through operations such as reflection, expansion, outer contraction, and inner contraction to find the global optimal solution. Specifically, the algorithm first starts from the midpoint position and performs a reflection operation. The reflection operation moves the midpoint position in the opposite direction by a certain distance to expand the search range. If the fitness value of the reflection point is better than the current optimal solution, an expansion operation is performed to further explore the potential in this direction; if the fitness value of the reflection point is poor, an outer contraction or inner contraction operation is performed to adjust the search direction and narrow the search range.

[0144] In addition, it should be noted that the third optimization equation refers to various formulas obtained by introducing the simplex method, including formulas for operations such as reflection, expansion, outer contraction, or inner contraction, where the reflection point x r position, as shown in Equation 10:

[0145] x r = x c + α(x c - x w )

[0146] (Equation 10)

[0147] where α is the reflection coefficient with a value of 1.

[0148] In addition, it should be noted that the fitness value of the reflection point is judged in three cases: First, if f r < f best , then continue to expand x r to obtain xe, as shown in Equation 11:

[0149] x e = x m + γ·(x r - x m )

[0150] (Equation 11)

[0151] where γ is the expansion coefficient with a value of 2. If f e < f best , it means the expansion is effective, and x e is used to replace x w ; otherwise, it means the expansion is ineffective, discard x e , and still use x r to replace x w .

[0152] Second, if f r > f best, the reflection direction is adjusted, and the position of the African vulture is contracted outward, so that the African vulture in a poor position is closer to the optimal position, and the outer contraction point x is obtained o , as shown in Equation 12:

[0153] x o = x m + β · (x w - x m )

[0154] (Equation 12)

[0155] where β is the contraction coefficient, and its value is 0.5. If f o < f best , it indicates that the outer contraction is effective, and x o is used to replace x w ; otherwise, it indicates that the outer contraction is ineffective, and x w is discarded, and x r is still used to replace x w .

[0156] Thirdly, if f best < f r < f worst , it indicates that the moving distance of the reflection point is too large, and inner contraction is required to obtain the inner contraction point x i , as shown in Equation 13:

[0157] x i = x m + β · (x r - x m )

[0158] (Equation 13)

[0159] where β is the contraction coefficient, and its value is 0.5. If f i < f best , it indicates that the inner contraction is effective, and x i is used to replace x w ; otherwise, it indicates that the inner contraction is ineffective, and x i is discarded, and x r is still used to replace x w .

[0160] It can be understood that after optimizing the African vulture individuals with poor fitness through the simplex method, the search can be prevented from falling into a stopped state and the search performance can be improved.

[0161] Step A40: Obtain the target African vulture optimization algorithm based on the African vulture optimization algorithm, the first optimization equation, the second optimization equation, and the third optimization equation.

[0162] It should be noted that in this step, the system initializes the population and sets the algorithm-related parameters, including the population size N, the spatial dimension dim, the search boundary [lb, ub], and the maximum number of iterations T max Then, according to the basic framework of the African vulture optimization algorithm, the fitness value of each individual in the population is calculated, and the best vulture and the second-best vulture are selected. These two vultures will serve as the guides in the subsequent iterative process and influence other vultures in the population. Then, it is judged whether the current algorithm running stage is the global exploration stage. If so, the African vulture optimization algorithm is optimized according to the first optimization equation (i.e., Equation 7). Specifically, in each iteration, the position of each vulture is perturbed by Cauchy-Gaussian mixed mutation. This perturbation strategy combines the characteristics of the Cauchy distribution and the Gaussian distribution, and can obtain a large step size through a large Cauchy mutation weight value at the initial stage of iteration, effectively avoiding the algorithm falling into a local optimal solution; while in the later stage of iteration, the weight value of Gaussian mutation is increased to generate a small step size for local perturbation of the current position, thus avoiding premature convergence to the local optimum and improving the convergence accuracy of the algorithm. After the global exploration stage, the algorithm enters the local exploitation stage. At this time, the position of the vulture is updated according to the second optimization equation (i.e., Equation 8). This equation introduces an adaptive weight coefficient strategy to distinguish the influence effects of the best vulture and the second-best vulture on the positions of other vultures. In the early stage of iteration, the guiding role of the second-best vulture is enhanced to improve the global search ability; while in the later stage of iteration, the guiding role of the best vulture is gradually enhanced to prompt the vulture to converge to the optimal solution as soon as possible, thereby improving the convergence speed of the algorithm and enhancing the local exploitation ability. In addition, during the entire running process of the algorithm, a third optimization equation (i.e., Equations 10-13) is introduced to further improve the search performance. This equation uses the simplex method to optimize the vulture individuals with poor fitness, and through operations such as reflection, expansion, outer contraction, and inner contraction, it avoids the algorithm falling into a stagnant state and improves the search performance. Combining the above three optimization equations, the target African vulture optimization algorithm (MY_AVOA) is obtained. On the basis of retaining the basic framework of the African vulture optimization algorithm, this algorithm significantly improves the convergence speed, global search ability, and convergence accuracy of the algorithm by introducing various optimization strategies such as Cauchy-Gaussian mixed mutation strategy, adaptive weight coefficient strategy, and simplex method

[0163] In this embodiment, in order to improve the optimization performance of the African vulture optimization algorithm, a multi-strategy improved algorithm is proposed. The Cauchy-Gaussian mixed mutation, adaptive weight coefficient, and simplex method are comprehensively used to improve the original algorithm, thereby improving the convergence speed and global search ability of the algorithm. And, according to the experimental results, the improved African vulture optimization algorithm (i.e., MY_AVOA) has significantly improved in terms of convergence speed, ability to jump out of the local optimum, and convergence accuracy

[0164] The present application also provides a voice classification device based on an improved African vulture. Please refer to Figure 10 , the voice classification device based on the improved African vulture includes:

[0165] A feature extraction module 10, configured to extract Mel-frequency cepstral coefficient features of voice data;

[0166] An algorithm optimization module 20, configured to optimize the African vulture optimization algorithm according to a preset optimization strategy to obtain a target African vulture optimization algorithm, where the preset optimization strategy includes a Cauchy Gaussian mixture mutation strategy, an adaptive weight coefficient strategy, and a simplex method strategy;

[0167] A model optimization module 30, configured to optimize hyperparameters of a long short-term memory network model according to the target African vulture optimization algorithm to obtain a target hyperparameter combination;

[0168] A voice classification module 40, configured to reconstruct the long short-term memory model according to the target hyperparameter combination and the Mel-frequency cepstral coefficient features to obtain a voice classification model for realizing voice classification.

[0169] The voice classification device based on the improved African vulture provided by the present application adopts the voice classification method based on the improved African vulture in the above embodiment, and can solve the technical problem of how to overcome the deficiencies of traditional voice classification algorithms and LSTM model parameter optimization and improve the accuracy and stability of urban voice classification. Compared with the prior art, the beneficial effects of the voice classification device based on the improved African vulture provided by the present application are the same as those of the voice classification method based on the improved African vulture provided by the above embodiment, and other technical features in the voice classification device based on the improved African vulture are the same as the features disclosed in the method of the above embodiment, and will not be elaborated herein.

[0170] The present application provides a voice classification device based on an improved African vulture. The voice classification device based on the improved African vulture includes: at least one processor; and a memory communicatively connected to the at least one processor; wherein, the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can execute the voice classification method based on the improved African vulture in the first embodiment above.

[0171] Next, refer to Figure 11, which shows a schematic structural diagram of a sound classification device based on an improved African vulture suitable for implementing the embodiments of the present application. The sound classification device based on the improved African vulture in the embodiments of the present application may include, but is not limited to, mobile terminals such as mobile phones, laptop computers, digital broadcast receivers, PDAs (Personal Digital Assistants), PADs (Portable Application Descriptions, tablet computers), PMPs (Portable Media Players), vehicle-mounted terminals (such as vehicle-mounted navigation terminals), etc., and fixed terminals such as digital TVs, desktop computers, etc. Figure 11 The shown sound classification device based on the improved African vulture is merely an example and should not impose any limitations on the functions and usage scope of the embodiments of the present application.

[0172] As Figure 11 shown, the sound classification device based on the improved African vulture may include a processing device 1001 (such as a central processing unit, a graphics processing unit, etc.), which may perform various appropriate actions and processes according to a program stored in a ROM (Read Only Memory) 1002 or a program loaded from a storage device 1003 into a RAM (Random Access Memory) 1004. In the RAM 1004, various programs and data required for the operation of the sound classification device based on the improved African vulture are also stored. The processing device 1001, the ROM 1002, and the RAM 1004 are connected to each other through a bus 1005. An input / output (I / O) interface 1006 is also connected to the bus. Generally, the following systems may be connected to the I / O interface 1006: an input device 1007 including, for example, a touch screen, a touchpad, a keyboard, a mouse, an image sensor, a microphone, an accelerometer, a gyroscope, etc.; an output device 1008 including, for example, an LCD (Liquid Crystal Display), a speaker, a vibrator, etc.; a storage device 1003 including, for example, a magnetic tape, a hard disk, etc.; and a communication device 1009. The communication device 1009 may allow the sound classification device based on the improved African vulture to communicate with other devices wirelessly or wireline to exchange data. Although the figure shows a sound classification device based on the improved African vulture having various systems, it should be understood that it is not required to implement or have all the shown systems. More or fewer systems may be alternatively implemented or had.

[0173] In particular, according to the embodiments disclosed in the present application, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, the embodiments disclosed in the present application include a computer program product, which includes a computer program carried on a computer-readable medium, and the computer program contains program codes for executing the methods shown in the flowcharts. In such an embodiment, the computer program can be downloaded and installed from a network through a communication device, or installed from a storage device 1003, or installed from a ROM 1002. When the computer program is executed by a processing device 1001, the above-mentioned functions defined in the methods of the embodiments disclosed in the present application are executed.

[0174] The sound classification device based on the improved African vulture provided by the present application adopts the sound classification method based on the improved African vulture in the above-mentioned embodiments, and can solve the technical problems of how to overcome the deficiencies of traditional sound classification algorithms and LSTM model parameter optimization, and improve the accuracy and stability of urban sound classification. Compared with the prior art, the beneficial effects of the sound classification device based on the improved African vulture provided by the present application are the same as those of the sound classification method based on the improved African vulture provided in the above-mentioned embodiments, and other technical features in the sound classification device based on the improved African vulture are the same as the features disclosed in the method of the previous embodiment, and will not be elaborated here.

[0175] It should be understood that the various parts disclosed in the present application can be implemented by hardware, software, firmware or a combination thereof. In the description of the above embodiments, specific features, structures, materials or characteristics can be combined in a suitable manner in any one or more embodiments or examples.

[0176] As mentioned above, the above are only specific embodiments of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art can easily think of changes or substitutions within the technical scope disclosed in the present application, and all should be covered by the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.

[0177] The present application provides a computer-readable storage medium, on which computer-readable program instructions (i.e., computer programs) are stored, and the computer-readable program instructions are used to execute the sound classification method based on the improved African vulture in the above-mentioned embodiments.

[0178] The computer-readable storage medium provided by this application can be, for example, a USB flash drive, but is not limited to electrical, magnetic, optical, electromagnetic, infrared, or semiconductor systems, devices, or components, or any combination of the above. More specific examples of computer-readable storage media may include, but are not limited to: electrical connections with one or more wires, portable computer disks, hard disks, RAM (Random Access Memory), ROM (Read Only Memory), EPROM (Erasable Programmable Read Only Memory or flash memory), optical fibers, CD-ROM (Compact Disc - Read Only Memory), optical storage devices, magnetic storage devices, or any suitable combination of the above. In this embodiment, the computer-readable storage medium can be any tangible medium that contains or stores a program, which can be used by or in conjunction with an instruction execution system, device, or component. The program code contained on the computer-readable storage medium can be transmitted using any appropriate medium, including but not limited to: wires, optical cables, RF (Radio Frequency), etc., or any suitable combination of the above.

[0179] The above computer-readable storage medium can be included in the sound classification device based on the improved African vulture; or it can exist separately and not be assembled into the sound classification device based on the improved African vulture.

[0180] The above computer-readable storage medium carries one or more programs. When the above one or more programs are executed by the sound classification device based on the improved African vulture, the sound classification device based on the improved African vulture is enabled to: extract the Mel-frequency cepstral coefficient features of the sound data; optimize the African vulture optimization algorithm according to a preset optimization strategy to obtain a target African vulture optimization algorithm, where the preset optimization strategy includes a Cauchy-Gaussian mixture mutation strategy, an adaptive weight coefficient strategy, and a simplex method strategy; optimize the hyperparameters of the long short-term memory network model according to the target African vulture optimization algorithm to obtain a target long short-term memory network model; and construct a sound classification model according to the target long short-term memory network model and the Mel-frequency cepstral coefficient features to achieve sound classification.

[0181] Computer program code for performing the operations of this application can be written in one or more programming languages or combinations thereof. The above-mentioned programming languages include object-oriented programming languages such as Java, Smalltalk, C++, and also include conventional procedural programming languages such as the "C" language or similar programming languages. The program code can be executed entirely on the user's computer, partially on the user's computer, executed as an independent software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In the case of a remote computer, the remote computer can be connected to the user's computer through any kind of network, including a LAN (Local Area Network) or a WAN (Wide Area Network), or it can be connected to an external computer (for example, by connecting through the Internet using an Internet service provider).

[0182] The flowcharts and block diagrams in the accompanying drawings illustrate the possible architectures, functions, and operations of systems, methods, and computer program products according to various embodiments of this application. In this regard, each block in the flowchart or block diagram can represent a module, a program segment, or a part of the code, and this module, program segment, or part of the code contains one or more executable instructions for implementing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the blocks may occur in a different order than that marked in the accompanying drawings. For example, two consecutively represented blocks can actually be executed substantially in parallel, and they can sometimes be executed in the reverse order, depending on the functions involved. It should also be noted that each block in the block diagram and / or flowchart, and the combination of blocks in the block diagram and / or flowchart, can be implemented by a dedicated hardware-based system for performing the specified functions or operations, or can be implemented by a combination of dedicated hardware and computer instructions.

[0183] The modules involved in the embodiments described in this application can be implemented in software or in hardware. Among them, the name of the module does not constitute a limitation on the unit itself in some cases.

[0184] The readable storage medium provided in this application is a computer-readable storage medium. The computer-readable storage medium stores computer-readable program instructions (i.e., computer programs) for performing the above-mentioned sound classification based on improved African vultures, and can solve the technical problems of how to overcome the deficiencies of traditional sound classification algorithms and LSTM model parameter optimization, and improve the accuracy and stability of urban sound classification. Compared with the prior art, the beneficial effects of the computer-readable storage medium provided in this application are the same as those of the sound classification based on improved African vultures provided in the above embodiments, and will not be elaborated here.

[0185] The present application also provides a computer program product, including a computer program, and when the computer program is executed by a processor, it implements the steps of the above-mentioned sound classification method based on the improved African vulture.

[0186] The computer program product provided by the present application can solve the technical problem of how to overcome the deficiencies of traditional sound classification algorithms and LSTM model parameter optimization and improve the accuracy and stability of urban sound classification. Compared with the prior art, the beneficial effects of the computer program product provided by the present application are the same as those of the sound classification method based on the improved African vulture provided in the above embodiments, and will not be elaborated herein.

[0187] The above are only partial embodiments of the present application, and thus do not limit the patent scope of the present application. Any equivalent structural transformation made under the technical concept of the present application by using the content of the specification and drawings of the present application, or any direct / indirect application in other related technical fields, is included in the patent protection scope of the present application.

Claims

1. A sound classification method based on improved African vultures, characterized in that: The method comprises: Extract the Mel-frequency cepstral coefficient features of sound data; Optimizing the African vulture optimization algorithm according to a preset optimization strategy to obtain a target African vulture optimization algorithm, wherein the preset optimization strategy includes a Cauchy-Gaussian mixed mutation strategy, an adaptive weight coefficient strategy, and a simplex method strategy; Optimizing the hyperparameters of the long short-term memory network model according to the target African vulture optimization algorithm to obtain a target hyperparameter combination; The long short-term memory model is reconstructed according to the target hyperparameter combination and the Mel-frequency cepstral coefficient features to obtain a sound classification model to achieve sound classification.

2. The method according to claim 1, characterized in that The step of optimizing the African vulture optimization algorithm according to the preset optimization strategy to obtain the target African vulture optimization algorithm comprises: Get the current hunger rate of the African vulture optimization algorithm; Determining the algorithm operation stage according to the range of the hunger rate; According to the algorithm running stage, the Cauchy-Gaussian mixed mutation strategy, the adaptive weight coefficient strategy and the simplex method strategy are called to optimize the African vulture optimization algorithm to obtain the target African vulture optimization algorithm.

3. The method according to claim 2, characterized in that The step of optimizing the African vulture optimization algorithm by calling the Cauchy-Gaussian mixture mutation strategy, the adaptive weight coefficient strategy and the simplex method strategy according to the algorithm running stage to obtain the target African vulture optimization algorithm comprises: When the algorithm operation phase is a global exploration phase, the African vulture optimization algorithm is optimized according to the Cauchy-Gaussian mixture mutation strategy to obtain a first optimization equation; When the algorithm operation stage is a local development stage, the African vulture optimization algorithm is optimized according to an adaptive weight coefficient strategy to obtain a second optimization equation; When the algorithm operation phase is a global optimization phase, the African vulture optimization algorithm is optimized according to the simplex method strategy to obtain a third optimization equation; A target African vulture optimization algorithm is obtained based on the African vulture optimization algorithm, the first optimization equation, the second optimization equation and the third optimization equation.

4. The method according to claim 3, characterized in that When the algorithm operation phase is a global exploration phase, the step of optimizing the African vulture optimization algorithm according to the Cauchy-Gaussian mixture mutation strategy to obtain a first optimization equation includes: When the algorithm operation phase is a global exploration phase, determining an adaptive weight and a greedy mechanism; The adaptive weight and the greedy mechanism are added to the African vulture optimization algorithm according to the Cauchy-Gauss mixture mutation strategy to obtain a first optimization equation.

5. The method according to claim 3, characterized in that When the algorithm operation stage is the local development stage, the step of optimizing the African vulture optimization algorithm according to the adaptive weight coefficient strategy to obtain the second optimization equation includes: When the algorithm operation stage is a local development stage, an adaptive weight coefficient is introduced according to an adaptive weight coefficient strategy; The adaptive weight coefficient is added to the African vulture optimization algorithm to obtain a second optimization equation.

6. The method according to claim 3, characterized in that When the algorithm operation stage is a global optimization stage, the step of optimizing the African vulture optimization algorithm according to the simplex method strategy to obtain a third optimization equation includes: When the algorithm operation stage is a global optimization stage, the population individuals of the African vulture optimization algorithm are sorted to obtain a first vulture position and a second vulture position; Obtaining a midpoint position according to the first vulture position and the second vulture position; A third optimization equation is obtained by performing reflection, expansion, external contraction or internal contraction operations based on the simplex method strategy and the midpoint position.

7. The method according to claim 1, characterized in that The step of optimizing the hyperparameters of the long short-term memory network model according to the target African vulture optimization algorithm to obtain a target hyperparameter combination includes: Setting hyperparameters of the long short-term memory network model, wherein the hyperparameters include the number of neurons in the first layer, the number of neurons in the second layer, the learning rate, and the drop rate; The hyperparameters are mapped to the target African vulture optimization algorithm, and the target African vulture optimization algorithm is iteratively optimized to obtain a target hyperparameter combination.

8. A sound classification device based on improved African vultures, characterized in that: The device comprises: A feature extraction module, used to extract Mel frequency cepstral coefficient features of sound data; An algorithm optimization module, used for optimizing the African vulture optimization algorithm according to a preset optimization strategy to obtain a target African vulture optimization algorithm, wherein the preset optimization strategy includes a Cauchy-Gaussian mixed mutation strategy, an adaptive weight coefficient strategy, and a simplex method strategy; A model optimization module, used to optimize the hyperparameters of the long short-term memory network model according to the target African vulture optimization algorithm to obtain a target hyperparameter combination; The sound classification module is used to reconstruct the long short-term memory model according to the target hyperparameter combination and the Mel-frequency cepstral coefficient characteristics to obtain a sound classification model to achieve sound classification.

9. A sound classification device based on improved African vultures, characterized in that: The device comprises: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the computer program is configured to implement the steps of the sound classification method based on improving African vultures according to any one of claims 1 to 7.

10. A storage medium, characterized in that: The storage medium is a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the steps of the sound classification method based on the improvement of African vultures are implemented as described in any one of claims 1 to 7.

Images