A voice signal detection method, device, apparatus and storage medium

Abstract

The application is suitable for the technical field of computers, and provides a voice signal detection method, device, equipment and storage medium, which comprises the following steps: performing filtering processing and frequency division operation on a to-be-detected audio frame to obtain a spectrum of the to-be-detected audio frame in a preset frequency range, obtaining frequency band energy and frequency band energy features of each frequency band in the spectrum, judging whether the to-be-detected audio frame is in a preset frame number range, determining the to-be-detected audio frame as a non-voice signal when the to-be-detected audio frame is in the frame number range, calculating total energy of the to-be-detected audio frame according to the frequency band energy when the to-be-detected audio frame is not in the frame number range, obtaining frequency band likelihood ratios of each frequency band according to the frequency band energy features, a pre-established voice signal model and a non-voice signal model, and judging whether the to-be-detected audio frame is a voice signal according to the total energy and the frequency band likelihood ratios, so that the accuracy of voice detection is improved.

Description

TECHNICAL FIELD

[0001] The present application belongs to the technical field of computer, and particularly relates to a voice signal detection method, device, equipment and storage medium. BACKGROUND

[0002] With the rapid development of artificial intelligence, people have increasingly realized the importance of speech recognition technology through practice, and the accuracy of voice signal detection directly determines whether the entire speech recognition system can achieve true intelligence as a key technology in the speech processing stage of the speech recognition system. The existing voice signal detection usually uses voice activity detection (Voice Activity Detection, abbreviated as VAD) technology to identify and eliminate long periods of silence in audio, saving bandwidth resources without reducing speech quality. However, the VAD algorithm based on zero-crossing rate and short-time energy cannot accurately detect the voice signal of the audio in a noisy environment, and the voice signal detection method based on deep neural network can accurately detect the audio in a high-noise environment, but the deep neural network has high algorithm complexity, large computation and high power consumption. SUMMARY

[0003] The purpose of the present application is to provide a voice signal detection method, device, equipment and storage medium, which aims to solve the problem of inaccurate voice signal detection due to the fact that the prior art cannot provide an effective voice signal detection method.

[0004] In one aspect, the present application provides a voice signal detection method, which comprises the following steps:

[0005] Filtering and frequency dividing the to-be-detected audio frame to obtain the spectrum of the to-be-detected audio frame in a preset frequency range, and obtaining the frequency band energy and frequency band energy feature of each frequency band in the spectrum;

[0006] Determining whether the to-be-detected audio frame is within a preset frame number range, and determining the to-be-detected audio frame as a non-speech signal when it is within the frame number range;

[0007] When the to-be-detected audio frame is not within the frame number range, calculating the total energy of the to-be-detected audio frame according to the frequency band energy of each frequency band, obtaining the frequency band likelihood ratio of each frequency band according to the frequency band energy feature, the pre-established voice signal model and the non-speech signal model, and determining whether the to-be-detected audio frame is a speech signal according to the total energy and the frequency band likelihood ratio.

[0008] In another aspect, the present application provides a voice signal detection device, which comprises:

[0009] a spectrum acquisition unit, configured to perform filtering processing and frequency division operation on the audio frame to be detected to obtain a spectrum of the audio frame to be detected in a preset frequency range, acquire frequency band energy and frequency band energy feature of each frequency band in the spectrum;

[0010] a range judgment unit, configured to judge whether the audio frame to be detected is in a preset frame number range, and determine the audio frame to be detected as a non-speech signal when the audio frame to be detected is in the frame number range;

[0011] a signal detection unit, configured to calculate total energy of the audio frame to be detected according to the frequency band energy of each frequency band when the audio frame to be detected is not in the frame number range, acquire frequency band likelihood ratio of each frequency band according to the frequency band energy feature, a pre-established speech signal model and the non-speech signal model, and judge whether the audio frame to be detected is a speech signal according to the total energy and the frequency band likelihood ratio.

[0012] In another aspect, the present application also provides a speech signal detection device, comprising a memory, a processor and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps of the method as described above when executing the computer program.

[0013] In another aspect, the present application also provides a computer readable storage medium, which stores a computer program, wherein the computer program is executable by a processor to implement the steps of the method as described above.

[0014] The present application performs filtering processing and frequency division operation on the audio frame to be detected to obtain a spectrum of the audio frame to be detected in a preset frequency range, acquires frequency band energy and frequency band energy feature of each frequency band in the spectrum, judges whether the audio frame to be detected is in a preset frame number range, determines the audio frame to be detected as a non-speech signal when the audio frame to be detected is in the frame number range, calculates total energy of the audio frame to be detected according to the frequency band energy of each frequency band when the audio frame to be detected is not in the frame number range, acquires frequency band likelihood ratio of each frequency band according to the frequency band energy feature, a pre-established speech signal model and the non-speech signal model, and judges whether the audio frame to be detected is a speech signal according to the total energy and the frequency band likelihood ratio, thereby improving the accuracy of speech detection. BRIEF DESCRIPTION OF DRAWINGS

[0015] Figure 1 is an implementation flowchart of the speech signal detection method provided by the first embodiment of the present application;

[0016] Figure 2 is an implementation flowchart of the speech signal detection method provided by the second embodiment of the present application;

[0017] Figure 3is a structural schematic diagram of a voice signal detection device provided by an embodiment three of the present application.

[0018] Figure 4 is a structural schematic diagram of a voice signal detection device provided by an embodiment four of the present application. DETAILED DESCRIPTION

[0019] In order to make the objects, technical solutions and advantages of the present application clearer, the present application is further described in detail below with reference to the drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and do not limit the present application.

[0020] The specific implementation of the present application is described in detail below with reference to specific embodiments:

[0021] Example One:

[0022] Figure 1 The implementation process of the voice signal detection method provided by the embodiment one of the present application is shown, for the convenience of description, only the part related to the embodiment of the present application is shown, and the details are as follows:

[0023] In step S101, the to-be-detected audio frame is filtered and frequency-division operated to obtain the spectrum of the to-be-detected audio frame in a preset frequency range, and the frequency band energy and the frequency band energy feature of each frequency band in the spectrum are obtained.

[0024] The embodiment of the present application is applicable to a voice signal detection device for voice signal detection of audio. Specifically, the voice signal detection of audio is realized by detecting each audio frame in the audio. The voice signal detection device can be a personal computer or a server. After receiving the to-be-detected audio frame, the voice signal detection device performs filtering processing on the to-be-detected audio frame to obtain a signal in a preset frequency range. As an example, signals below 80 Hz and above 4000 Hz can be filtered out. In this way, the signal after filtering operation does not contain direct current and alternating current interference signals below 80 Hz, nor high frequency signals above 4000 Hz. Then, frequency division operation is performed on the to-be-detected audio frame after filtering processing, so as to obtain the spectrum of the filtered to-be-detected audio frame. Finally, the spectrum feature of the spectrum, the frequency band energy and the frequency band energy feature of each frequency band in the spectrum are obtained. The spectrum feature includes the energy distribution of the spectrum of the to-be-detected audio frame, i.e. the energy size at each frequency. The frequency band energy refers to the total energy in a frequency band, and the frequency band energy feature includes the energy size at each frequency in the frequency band.

[0025] In step S102, it is judged whether the to-be-detected audio frame is within a preset frame number range. When it is within the frame number range, the to-be-detected audio frame is determined as a non-voice signal.

[0026] In the embodiment of the present application, considering that the beginning part of the recorded audio is usually only environmental noise, when determining whether the audio frame to be detected is a non-speech signal, it can be determined whether the audio frame to be detected is within a preset frame number range, and when it is within the frame number range, the audio frame to be detected is directly determined as a non-speech signal, thereby simplifying the judgment process of the non-speech signal. Specifically, the sequence number of the audio frame to be detected, i.e., the order of the audio frame to be detected in all audio frames of the input audio, can be obtained, and then it is determined according to the sequence number whether the audio frame to be detected is within the preset frame number range, and when it is within the frame number range, the audio frame to be detected is determined as a non-speech signal. For example, when the audio frame to be detected is the 5th frame to be detected in the audio to be detected, if the preset frame number is 8, the audio to be detected is within the preset frame number range, and the audio frame is determined as a non-speech signal.

[0027] In a preferred embodiment, when the audio frame to be detected is a non-speech signal, the pre-established non-speech signal model is updated according to the spectral feature of the current audio frame to be detected, so that the non-speech signal model is updated in real time by the current detected audio, to improve the effectiveness of the non-speech signal model in subsequent use, wherein the non-speech signal model is used to extract the energy feature of the non-speech audio frame.

[0028] In step S103, when the audio frame to be detected is not within the frame number range, the total energy of the audio frame to be detected is calculated according to the frequency band energy of each frequency band, the frequency band likelihood ratio of each frequency band is obtained according to the frequency band energy feature, the pre-established speech signal model and the non-speech signal model, and whether the audio frame to be detected is a speech signal is determined according to the total energy and the frequency band likelihood ratio.

[0029] In the embodiment of the present application, the speech signal model is used to extract the energy feature of the speech audio frame. When the audio frame to be detected is not within the preset frame number range, the total energy of the current audio frame to be detected is calculated according to the frequency band energy of each frequency band, and the frequency band likelihood ratio of each frequency band is further obtained according to the frequency band energy feature of each frequency band and the pre-established speech signal model and non-speech signal model, the frequency band likelihood ratio indicating the ratio of the speech component to the non-speech component contained in each frequency band of the current audio frame. After obtaining the total energy of the audio frame to be detected and the frequency band likelihood ratio of each frequency band, it is determined according to the total energy and the frequency band likelihood ratio whether the current audio frame to be detected is a speech signal.

[0030] In the embodiment of the present application, after receiving the audio to be detected, the audio frame to be detected is filtered and frequency-division operated to obtain the frequency spectrum of the audio frame to be detected in a preset frequency range, the frequency band energy of each frequency band in the frequency spectrum and the frequency band energy feature are obtained, it is judged whether the audio frame to be detected is in a preset frame number range, when in the frame number range, the audio frame to be detected is determined as a non-speech signal, when the audio frame to be detected is not in the frame number range, the total energy of the audio frame to be detected is calculated according to the frequency band energy of each frequency band, the frequency band likelihood ratio of each frequency band is obtained according to the frequency band energy feature, the pre-established speech signal model and the non-speech signal model, and whether the audio frame to be detected is a speech signal is judged according to the total energy and the frequency band likelihood ratio, thereby improving the accuracy of speech detection.

[0031] Example Two:

[0032] Figure 2 The implementation process of the speech signal detection method provided by the second embodiment of the present application is shown, only the part related to the embodiment of the present application is shown for the convenience of description, and the details are described as follows:

[0033] In step S201, the audio frame to be detected is filtered and frequency-division operated to obtain the frequency spectrum of the audio frame to be detected in a preset frequency range, and the frequency band energy of each frequency band in the frequency spectrum and the frequency band energy feature are obtained.

[0034] In step S202, it is judged whether the audio frame to be detected is in a preset frame number range, when in the frame number range, the audio frame to be detected is determined as a non-speech signal.

[0035] In the embodiment of the present application, the implementation manners of steps S201-S202 are the same as those of steps S101-S102 in the first embodiment, and the corresponding description of the first embodiment can be referred to, and details are not repeated here.

[0036] In step S203, when the audio frame to be detected is not in the frame number range, the total energy of the audio frame to be detected is calculated according to the frequency band energy of each frequency band, the frequency band likelihood ratio of each frequency band is obtained according to the frequency band energy feature, the pre-established speech signal model and the non-speech signal model, and whether the audio frame to be detected is a speech signal is judged according to the total energy and the frequency band likelihood ratio.

[0037] In the embodiment of the present application, when the audio frame to be detected is not within the preset frame number range, the total energy of the current audio frame to be detected is calculated according to the frequency band energy of each frequency band, and the frequency band likelihood ratio of each frequency band is further obtained according to the frequency band energy feature of each frequency band, the pre-established speech signal model and non-speech signal model. After the total energy of the audio frame to be detected and the frequency band likelihood ratio of each frequency band are obtained, whether the current audio frame to be detected is a speech signal is determined according to the total energy and the frequency band likelihood ratio, so that the speech signal model and the non-speech signal model are used to realize accurate detection of the audio frame in a strong noise environment. Specifically, when the total energy of the audio frame to be detected is calculated according to the frequency band energy of each frequency band, the frequency band energy of each frequency band is added, and the obtained result is set as the total energy of the audio frame to be detected.

[0038] In a specific embodiment, when the frequency band likelihood ratio of each frequency band is obtained according to the frequency band energy feature, the pre-established speech signal model and non-speech signal model, the speech energy feature of the speech audio frame is obtained in advance through the speech signal model, the non-speech energy feature of the non-speech audio frame is obtained in advance through the non-speech signal model, the similarity between the frequency band energy feature of each frequency band in the audio frame to be detected and the speech energy feature is obtained, which is referred to as speech similarity for convenience of description, and the speech similarity represents the probability that the current frequency band has the speech energy feature. The similarity between the frequency band energy feature and the non-speech energy feature is further obtained, which is referred to as non-speech similarity for convenience of description, and the non-speech similarity represents the probability that the current frequency band has the non-speech energy feature. The ratio of the speech similarity and the non-speech similarity is taken as the likelihood ratio of the current frequency band.

[0039] Preferably, the speech signal model and the non-speech signal model are Gaussian mixture models, so as to accurately quantify the speech signal and the non-speech signal, and further accurately obtain the related features of the speech signal and the non-speech signal.

[0040] Preferably, when whether the audio frame to be detected is a speech signal is determined according to the total energy and the frequency band likelihood ratio, firstly, whether the total energy of the audio frame to be detected is greater than a preset speech energy threshold is determined. If the total energy is greater than the speech energy threshold, the current audio frame to be detected is determined as a speech signal. If the total energy is not greater than the speech energy threshold, the frequency band likelihood ratio of each frequency band is determined. Whether the frequency band likelihood ratio of each frequency band is greater than a preset likelihood ratio threshold is determined. If the frequency band likelihood ratio is greater than the likelihood ratio threshold, the current audio frame to be detected is determined as a speech signal. If the frequency band likelihood ratio is not greater than the likelihood ratio threshold, the current audio frame to be detected is determined as a non-speech signal. Therefore, the audio frame to be detected is determined in multiple levels through the energy and the frequency band likelihood ratio, so as to improve the accuracy and success rate of the determination.

[0041] In a specific embodiment, when judging whether the band likelihood ratio is greater than the preset likelihood ratio threshold, if the band likelihood ratio is greater than the likelihood ratio threshold, the band likelihood ratio of each band is compared with the likelihood ratio threshold respectively, if the band likelihood ratio of at least one band is greater than the likelihood ratio threshold, the audio frame to be detected is determined as a speech signal, thereby improving the accuracy of speech signal detection.

[0042] In step S204, when the audio frame to be detected is a speech signal, the speech signal model is updated according to the spectral feature of the audio frame to be detected.

[0043] In the embodiment of the present application, when the current audio frame to be detected is determined as a speech signal after multiple judgments, the speech signal model is updated according to the spectral feature of the audio frame to be detected, so that the parameters of the speech signal model can be updated in time, thereby ensuring the effectiveness of the model.

[0044] In step S205, when the audio frame to be detected is a non-speech signal, the non-speech signal model is updated according to the spectral feature of the audio frame to be detected.

[0045] In the embodiment of the present application, when the current audio frame to be detected is determined as a non-speech signal after multiple judgments, the non-speech signal model is updated according to the spectral feature of the audio frame to be detected, so that the parameters of the non-speech signal model can be updated in time, thereby ensuring the effectiveness of the model.

[0046] In the embodiment of the present application, after receiving the audio to be detected, the audio frame to be detected is filtered and frequency-divided to obtain the spectrum of the audio frame to be detected in a preset frequency range, the band energy and the band energy feature of each band in the spectrum are obtained, it is judged whether the audio frame to be detected is within a preset frame number range, when it is within the frame number range, the audio frame to be detected is determined as a non-speech signal, when it is not within the frame number range, the total energy of the audio frame to be detected is calculated according to the band energy of each band, the band likelihood ratio of each band is obtained according to the band energy feature, the pre-established speech signal model and the non-speech signal model, it is judged whether the audio frame to be detected is a speech signal according to the total energy and the band likelihood ratio, when the audio frame to be detected is a speech signal, the speech signal model is updated according to the spectral feature of the audio frame to be detected, when the audio frame to be detected is a non-speech signal, the non-speech signal model is updated according to the spectral feature of the audio frame to be detected, thereby updating the parameters of the speech signal model and the non-speech signal model in time, improving the effectiveness and accuracy of speech signal detection.

[0047] Example Three:

[0048] Figure 3The structure of the speech signal detection device provided by the third embodiment of the present application is shown. For the convenience of description, only the parts related to the embodiments of the present application are shown, which include:

[0049] The spectrum obtaining unit 31 is configured to perform filtering processing and frequency division operation on the audio frame to be detected to obtain the spectrum of the audio frame to be detected in a preset frequency range, and obtain the frequency band energy and the frequency band energy feature of each frequency band in the spectrum.

[0050] The range judging unit 32 is configured to judge whether the audio frame to be detected is in a preset frame number range. When the audio frame to be detected is in the frame number range, the audio frame to be detected is determined as a non-speech signal.

[0051] The signal detecting unit 33 is configured to, when the audio frame to be detected is not in the frame number range, calculate the total energy of the audio frame to be detected according to the frequency band energy of each frequency band, obtain the frequency band likelihood ratio of each frequency band according to the frequency band energy feature, the pre-established speech signal model and the non-speech signal model, and judge whether the audio frame to be detected is a speech signal according to the total energy and the frequency band likelihood ratio.

[0052] In the embodiments of the present application, each unit of the speech signal detection device can be realized by a corresponding hardware or software unit. Each unit can be an independent software or hardware unit, or can be integrated into a software or hardware unit. The specific implementation of each unit can refer to the description of the foregoing method embodiments, which will not be described here.

[0053] Example Four:

[0054] Figure 4 The structure of the speech signal detection device provided by the fourth embodiment of the present application is shown. For the convenience of description, only the parts related to the embodiments of the present application are shown.

[0055] The computer device 4 of the embodiment of the present application includes a processor 40, a memory 41, and a computer program 42 stored in the memory 41 and executable on the processor 40. The processor 40 implements the steps in the speech signal detection method embodiments described above when executing the computer program 42, such as the steps S101 to S103 shown. Figure 1 Alternatively, the processor 40 implements the functions of each unit in the foregoing device embodiments when executing the computer program 42, such as the functions of the units 31 to 33 shown. Figure 3

[0056] ​In the embodiment of the present application, after receiving the audio to be detected, the audio frame to be detected is subjected to filtering processing and frequency division operation to obtain the frequency spectrum of the audio frame to be detected in a preset frequency range, the frequency band energy and the frequency band energy feature of each frequency band in the frequency spectrum are obtained, it is judged whether the audio frame to be detected is in a preset frame number range, when in the frame number range, the audio frame to be detected is determined as a non-speech signal, when the audio frame to be detected is not in the frame number range, the total energy of the audio frame to be detected is calculated according to the frequency band energy of each frequency band, the frequency band likelihood ratio of each frequency band is obtained according to the frequency band energy feature, the pre-established speech signal model and the non-speech signal model, and it is judged whether the audio frame to be detected is a speech signal according to the total energy and the frequency band likelihood ratio, thereby improving the accuracy of speech detection.

[0057] The steps implemented by the processor 40 in the recommendation device 4 when implementing the speech signal detection method by executing the computer program 42 can refer to the descriptions of the foregoing method embodiments, and will not be described herein again.

[0058] Example Five:

[0059] In the embodiment of the present application, a computer readable storage medium is provided, which stores a computer program, and the computer program implements the steps in the speech signal detection method embodiments described above when executed by a processor, for example, the steps S101 to S103 shown in the foregoing method embodiments. Figure 1 Or, the computer program implements the functions of the units in the foregoing device embodiments when executed by a processor, for example, the functions of the units 31 to 33 shown in the foregoing device embodiments. Figure 3 Or, the computer program implements the functions of the units in the foregoing device embodiments when executed by a processor, for example, the functions of the units 31 to 33 shown in the foregoing device embodiments.

[0060] In the embodiment of the present application, after receiving the audio to be detected, the audio frame to be detected is subjected to filtering processing and frequency division operation to obtain the frequency spectrum of the audio frame to be detected in a preset frequency range, the frequency band energy and the frequency band energy feature of each frequency band in the frequency spectrum are obtained, it is judged whether the audio frame to be detected is in a preset frame number range, when in the frame number range, the audio frame to be detected is determined as a non-speech signal, when the audio frame to be detected is not in the frame number range, the total energy of the audio frame to be detected is calculated according to the frequency band energy of each frequency band, the frequency band likelihood ratio of each frequency band is obtained according to the frequency band energy feature, the pre-established speech signal model and the non-speech signal model, and it is judged whether the audio frame to be detected is a speech signal according to the total energy and the frequency band likelihood ratio, thereby improving the accuracy of speech detection.

[0061] The computer readable storage medium of the embodiment of the present application can include any entity or device, recording medium capable of carrying computer program code, for example, ROM / RAM, magnetic disk, optical disk, flash memory and the like memory.

[0062] The above merely describes preferred embodiments of the present application, and is not used to limit the present application, any modification, equivalent replacement and improvement within the spirit and principle of the present application should be included in the protection scope of the present application.

Claims

1. A voice signal detection method characterized by, The method comprises the following steps: filtering and frequency dividing the audio frame to be detected to obtain a spectrum of the audio frame to be detected in a preset frequency range, and obtaining frequency band energies of each frequency band in the spectrum and a frequency band energy feature; determining whether the audio frame to be detected is within a preset frame number range, and determining the audio frame to be detected as a non-speech signal when the audio frame to be detected is within the frame number range; when the audio frame to be detected is not within the frame number range, calculating a total energy of the audio frame to be detected according to the frequency band energies of each frequency band, obtaining a frequency band likelihood ratio of each frequency band according to the frequency band energy feature, a pre-established speech signal model and a non-speech signal model, and determining whether the audio frame to be detected is a speech signal according to the total energy and the frequency band likelihood ratio, the step of determining whether the audio frame to be detected is a speech signal according to the total energy and the frequency band likelihood ratio comprises: determining whether the total energy is greater than a preset speech energy threshold, and determining the audio frame to be detected as a speech signal when the total energy is greater than the speech energy threshold; determining whether the frequency band likelihood ratio is greater than a preset likelihood ratio threshold when the total energy is not greater than the speech energy threshold, and determining the audio frame to be detected as a speech signal when the frequency band likelihood ratio is greater than the likelihood ratio threshold; determining the audio frame to be detected as a non-speech signal when the frequency band likelihood ratio is not greater than the likelihood ratio threshold.

2. The method of claim 1, wherein, the step of determining whether the audio frame to be detected is a speech signal according to the total energy and the frequency band likelihood ratio is followed by: updating the speech signal model according to a spectrum feature of the audio frame to be detected when the audio frame to be detected is a speech signal; updating the non-speech signal model according to a spectrum feature of the audio frame to be detected when the audio frame to be detected is a non-speech signal.

3. The method of claim 1, wherein, the step of determining whether the frequency band likelihood ratio is greater than a preset likelihood ratio threshold, and determining the audio frame to be detected as a speech signal when the frequency band likelihood ratio is greater than the likelihood ratio threshold, comprises: comparing the frequency band likelihood ratio of each frequency band with the likelihood ratio threshold respectively, and determining the audio frame to be detected as a speech signal when the frequency band likelihood ratio of at least one frequency band is greater than the likelihood ratio threshold.

4. The method of claim 1, wherein, the step of calculating the total energy of the audio frame to be detected according to the frequency band energies of each frequency band comprises: adding the frequency band energies of each frequency band together, and setting the obtained result as the total energy of the audio frame to be detected.

5. The method of claim 1, wherein, the step of obtaining the frequency band likelihood ratio of each frequency band according to the frequency band energy feature, a pre-established speech signal model and a non-speech signal model comprises: obtaining a speech energy feature of a speech signal through the speech signal model, obtaining a non-speech energy feature of a non-speech signal through the non-speech signal model, obtaining a speech similarity between the frequency band energy feature and the speech energy feature and a non-speech similarity between the frequency band energy feature and the non-speech energy feature, and the ratio of the speech similarity to the non-speech similarity being the frequency band likelihood ratio.

6. The method of claim 1, wherein, the speech signal model and the non-speech signal model are a mixture Gaussian model.

7. A voice signal detecting apparatus characterized by comprising: the device comprises: The spectrum acquisition unit is configured to perform filtering processing and frequency division operation on the audio frame to be detected to obtain a spectrum of the audio frame to be detected in a preset frequency range, and acquire a frequency band energy and a frequency band energy feature of each frequency band in the spectrum. The range judgment unit is configured to judge whether the audio frame to be detected is in a preset frame number range, and determine the audio frame to be detected as a non-speech signal when the audio frame to be detected is in the frame number range. The signal detection unit is configured to, when the audio frame to be detected is not in the frame number range, calculate a total energy of the audio frame to be detected according to the frequency band energy of each frequency band, acquire a frequency band likelihood ratio of each frequency band according to the frequency band energy feature, a pre-established speech signal model and the non-speech signal model, and judge whether the audio frame to be detected is a speech signal according to the total energy and the frequency band likelihood ratio. The step of judging whether the audio frame to be detected is a speech signal according to the total energy and the frequency band likelihood ratio comprises: judging whether the total energy is greater than a preset speech energy threshold, and determining the audio frame to be detected as a speech signal when the total energy is greater than the speech energy threshold; judging whether the frequency band likelihood ratio is greater than a preset likelihood ratio threshold when the total energy is not greater than the speech energy threshold, and determining the audio frame to be detected as a speech signal when the frequency band likelihood ratio is greater than the likelihood ratio threshold; and determining the audio frame to be detected as a non-speech signal when the frequency band likelihood ratio is not greater than the likelihood ratio threshold.

8. A voice signal detection device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, The processor executes the computer program to implement the steps of the method according to any one of claims 1 to 6.

9. A computer-readable storage medium storing a computer program, the computer program comprising instructions that, when executed by a computer, cause the computer to perform the method of any one of claims 1 to 8. The computer program is executed by the processor to implement the steps of the method according to any one of claims 1 to 6.

Images