Noise processing method and system based on artificial intelligence

Abstract

The invention provides a noise processing method and system based on artificial intelligence, and relates to the technical field of noise processing. The method comprises the following steps: acquiring historical voice noise processing data, and performing noise feature extraction based on an environment background to form background noise feature data; according to the background noise feature data, performing denoising processing on the to-be-processed voice information to form initial denoised voice information; and noise screening analysis based on smoothness and semantics is carried out on the initial de-noised voice information, and target voice information is extracted. According to the method, the feature data of the environmental background noise is established by using the big data, so that the noise in the recorded voice can be identified and most of the noise can be screened out more effectively, and clear target voice information can be obtained.

Description

Technical Field

[0001] The present invention relates to the technical field of noise processing, and in particular to a noise processing method and system based on artificial intelligence. Background Art

[0002] Voice collection is to extract the target voice from the complex background sound. Due to the presence of background sound, there will be a lot of noise mixed in, making the original recorded voice more mixed, and it is not possible to clearly obtain the information of the target voice. Therefore, the original recorded voice needs to be processed by noise reduction to clearly highlight the target voice.

[0003] Currently, most speech noise reduction processing is performed directly based on the large sound feature gap between the ambient background sound and the target speech. However, when this gap is relatively small, the noise reduction effect is no longer so effective, so that the noise after noise reduction still exists in the speech that affects the acquisition of the target speech, and no obvious noise reduction effect can be achieved.

[0004] Therefore, designing a noise processing method and system based on artificial intelligence, by using big data to establish the characteristic data of environmental background noise, so as to more effectively identify and filter out most of the noise in the recorded voice, and ensure that clear and distinct target voice information can be obtained, is an urgent problem to be solved. Summary of the invention

[0005] The purpose of the present invention is to provide a noise processing method based on artificial intelligence. By extracting feature information of background noise in different environments from historical speech noise processing big data, the background noise of the processed speech can be screened out by using feature data of big data. Since the extracted feature data comes from big data, the noise information loss existing in a single denoising feature can be avoided to a certain extent, so that the noise feature data in different environments can basically cover most of the noise in the corresponding environmental scenes. At the same time, on the basis of removing most of the environmental noise, the smoothing processing and semantic analysis of the speech information are used to realize the deep extraction of different types of target speech information, and the target speech information is completely extracted from a certain small amount of noise information, so as to ensure the accuracy of the extracted target speech information, make the formed speech information clearer, and effectively avoid the influence of noise on its information transmission.

[0006] The purpose of the present invention is also to provide a noise processing system based on artificial intelligence. The system is configured to collect environmental background noise feature information using historical speech noise processing big data to achieve environmental background noise screening of speech information to be processed, and accurately and reasonably extract the target speech information as a whole, efficiently and accurately achieving more effective noise reduction processing, which is an important material basis for achieving noise reduction processing of the speech to be processed.

[0007] In a first aspect, the present invention provides a noise processing method based on artificial intelligence, comprising acquiring historical speech noise processing data, performing noise feature extraction based on environmental background, and forming background noise feature data; performing denoising processing on speech information to be processed according to the background noise feature data, and forming initial denoised speech information; performing smoothing and semantic-based noise screening analysis on the initial denoised speech information, and extracting target speech information.

[0008] In the present invention, the method extracts feature information of background noise in different environments from historical speech noise processing big data, and realizes the use of big data feature data to screen out the background noise of the processed speech. Since the extracted feature data comes from big data, the noise information loss existing in a single denoising feature can be avoided to a certain extent, so that the noise feature data in different environments can basically cover most of the noise in the corresponding environmental scenes. At the same time, on the basis of removing most of the environmental noise, the smoothing processing and semantic analysis of the speech information are used to realize the deep extraction of different types of target speech information, and it is completely extracted from a certain small amount of noise information, so as to ensure the accuracy of the extracted target speech information, make the formed speech information clearer, and effectively avoid the influence of noise on its information transmission.

[0009] As a possible implementation method, historical speech noise processing data is obtained, and noise features are extracted based on the environmental background to form background noise feature data, including: performing volume analysis on different historical speech processing information in the historical speech noise processing data for historical speech target information to form noise volume differentiation feature data; performing background noise feature extraction on different historical speech processing information under different environmental conditions within the volume range of the historical speech target information to form target volume range background noise feature data.

[0010] In the present invention, the environmental background noise feature information is extracted from the historical voice noise processing data, and two aspects are mainly considered. The first aspect is that the environmental noise contained in the voice information to be processed itself, which is initially targeted in the historical noise removal process, will have a relatively large volume of noise in the feature information of the sound, and there is a large gap between the sound features of the target voice information to be extracted, so it can be screened in sequence. Since this obvious feature information gap is often included in the collected voice data to be processed, the historical noise feature data can be used to fully extract this feature information, so as to quickly achieve the removal of the noise with a large gap between the feature information and the feature information of the target voice information. Here, the feature information of the noise with a large gap is mainly considered in the volume of the noise. It can be understood that the target voice information to be collected must be clearly and completely collected during the collection process, and the clear and complete collection requires a certain range of volume size, so by analyzing the volume size, the noise covered by the volume that does not correspond to the volume range of the target voice information can be screened out. The second aspect is that the noise forms in different background environments are different. Not only are there great differences in non-steady-state noise, but there are also great differences in steady-state noise. Therefore, the extraction of noise feature information based on different environmental backgrounds can ensure that the extracted noise feature data is more targeted, so that the subsequent noise reduction processing can more fully achieve noise screening after reference.

[0011] As a possible implementation method, volume analysis is performed on different historical speech processing information in the historical speech noise processing data for historical speech target information to form noise volume distinguishing feature data, including: determining the volume range of historical speech target information corresponding to different historical speech processing information in the historical speech noise processing data to form a historical target speech volume range corresponding to different historical speech processing information; and performing a AND operation on all historical target speech volume ranges to form a noise volume distinguishing feature range.

[0012] In the present invention, for the collection of noise feature information that has obvious sound feature information differences with the target voice information, considering the coverage of these noise feature data as much as possible under big data, and then for the subsequent effective removal of this part of the noise, after obtaining the noise feature information with obvious feature differences in the corresponding voice information to be processed based on the volume range of the historical voice target information, all the obtained noise feature information is combined and calculated to ensure that all this type of noise feature information in the historical data can be covered.

[0013] As a possible implementation method, background noise features under different environmental conditions within the volume range of historical voice target information are extracted for different historical voice processing information to form background noise feature data in a target volume range, including: extracting corresponding voice information within the volume range of historical voice target information for different historical voice processing information to form historical voice information in a corresponding volume range; clustering historical voice information in different volume ranges based on different background environment types to form a historical voice information set of an environment type and volume range; and performing a union operation on sound wave information of all volume range historical voice information in the historical voice information set of different environment type and volume range to form background noise feature data in a target volume range of the corresponding background environment.

[0014] In the present invention, for noise with the same volume range as the target voice information, the remaining noise information is obtained by using the historical voice target information that has been calibrated in the historical voice processing information. Since there is a large gap in the noise feature information generated by different environmental backgrounds, the noise information obtained is clustered only under the same background environment. The definition of the background environment can be carried out according to actual needs. The clustering of noise feature information based on different environmental backgrounds can further improve the pertinence of the noise feature information and improve the accuracy of the subsequent noise reduction processing of the voice information to be processed. It should be noted that for this type of noise feature information, the sound wave information is used as the aspect of obtaining the feature information, so that the acquired feature information can be guaranteed to be multi-dimensional, such as the frequency, period interval, amplitude, etc. that can be obtained from the sound wave information. The comprehensive feature information in many aspects is more recognizable, making the subsequent noise reduction using the feature information more accurate.

[0015] As a possible implementation method, based on the background noise feature data, the voice information to be processed is denoised to form initial denoised voice information, including: based on the noise volume distinction feature range, voice information within the noise volume distinction feature range is extracted to form target processing interval voice information; based on different target volume interval background noise feature data, background noise is denoised on the voice information in the target processing interval to form initial denoised voice information.

[0016] In the present invention, after the noise feature information extraction based on the historical speech processing information is completed, the extracted noise feature information can be used to perform noise reduction processing on the speech information to be processed. According to the form of the extracted background noise feature data, the noise reduction of the speech information to be processed includes two sequential steps. First, the noise volume is used to distinguish the feature range to perform noise reduction based on the volume feature range on the speech information to be processed, that is, to remove the noise part with a large difference between the feature information and the feature information of the target language information. Secondly, the target volume interval background noise feature data is used to remove most of the environmental background noise of the speech information in the target processing interval. Of course, it is necessary to determine the environmental background of the speech information in the target processing interval before the removal, and then extract the corresponding target volume interval background noise feature data for noise reduction.

[0017] As a possible implementation method, according to different target volume interval background noise feature data, the target processing interval voice information is subjected to background noise denoising processing to form initial denoised voice information, including: extracting the sound wave information of the target processing interval voice information, and matching and analyzing it with different target volume interval background noise feature data, determining the target volume interval background noise feature data with the largest number of matched sound waves, and using the background environment corresponding to the target volume interval background noise feature data as the background environment corresponding to the target processing interval voice information; removing the voice information corresponding to all the sound wave information matching the target volume interval background noise feature data in the target processing interval voice information to form initial denoised voice information.

[0018] In the present invention, the background noise of the target processing interval voice information is denoised using the corresponding target volume interval background noise feature data. The specific processing method is to determine whether there is noise feature information in the target processing interval voice information that matches the target volume interval background noise feature data. Only when there is a match can it be determined that the matched voice information belongs to environmental noise. This also avoids the loss of target voice information due to the removal of target voice information to a certain extent.

[0019] As a possible implementation method, the initial denoised speech information is subjected to a smoothing and semantic-based noise screening analysis to extract the target speech information, including: performing smoothing noise screening on the initial denoised speech information to determine the smoothed speech information to be confirmed, and performing a judgment analysis on the smoothed speech information to be confirmed based on music characteristics to form smooth speech type calibration information; performing semantic noise screening analysis on the initial denoised speech information to form the target language speech information.

[0020] In the present invention, after the processed voice information is denoised by using background noise feature data, it is considered that the remaining voice information may still have some new noise information such as new noise information that has not been covered by big data and noise whose feature information is close to the target voice feature information, so it is necessary to perform reasonable screening to extract the target voice information. Here, the target voice information is extracted from two aspects. On the one hand, if the target voice information belongs to a voice with obvious regularity, such as different types of music, then since this regular voice is usually smoothed, the smoothness is higher than that of the noise, so the analysis based on the smoothness can extract the voice with obvious regularity. On the other hand, if the target voice is language information, then the target voice has obvious semantic information compared to the noise, and at the same time, the language information that can be obtained by the human voice mixed in the noise is more, so the extraction of this type of target voice information can be realized by analyzing the voice information.

[0021] As a possible implementation method, the initial denoised speech information is subjected to smooth noise screening to determine the smooth speech information to be confirmed, and the smooth speech information to be confirmed is subjected to judgment analysis based on music characteristics to form smooth speech type calibration information, including: extracting the spectrum of all speech in the initial denoised speech information to form the original speech spectrum of the corresponding speech; derivativeizing the spectrum lines in the original speech spectrum to form the corresponding spectrum line derivative information; determining the speech information corresponding to the spectrum line with the least number of non-differentiable position points in the derivative information of different spectrum lines as the smooth speech information to be confirmed; setting a frequency change threshold, and performing the following analysis on the smooth speech information to be confirmed: if the maximum frequency difference of the spectrum lines corresponding to the smooth speech information to be confirmed does not exceed the frequency change threshold, the smooth speech information to be confirmed is calibrated as non-music smooth speech information; if the maximum frequency difference of the spectrum lines corresponding to the smooth speech information to be confirmed exceeds the frequency change threshold, the smooth speech information to be confirmed is calibrated as music smooth speech information.

[0022] In the present invention, the extraction of target speech information with regularity is mainly achieved by collecting the spectrum of all speech in the initial denoised speech information for smoothness analysis. If it is a target speech with regularity, due to its high smoothness, the derivative information formed after the spectrum is derivatived will basically not have many non-differentiable position points. Therefore, the speech information with the least statistical number obtained by counting the number of non-differentiable position points is most likely to be the target speech information with regularity. Of course, to determine whether there is such obvious regularity, it is also necessary to judge the range of frequency change. After all, there is an obvious difference in the range of frequency between single tone and concert.

[0023] As a possible implementation method, semantic noise screening analysis is performed on the initial denoised speech information to form target language speech information, including: extracting language phrases from different speech information in the initial denoised speech information to form a corresponding speech phrase set; and determining the speech information with the largest number of language phrases in all speech phrase sets as the target language speech information.

[0024] In the present invention, the noise screening of semantic analysis of the initial denoised speech information is mainly to extract the language phrases conveyed by different speech information. Considering that if the target speech is a human voice, the phrases are coherent and obvious, the target speech information can be determined by the size of the phrase quantity.

[0025] In a second aspect, the present invention provides an artificial intelligence-based noise processing system, which is configured to: obtain historical speech noise processing data, perform noise feature extraction based on the environmental background, and form background noise feature data; perform denoising processing on the speech information to be processed based on the background noise feature data to form initial denoised speech information; perform smoothing and semantic-based noise screening analysis on the initial denoised speech information to extract the target speech information.

[0026] In the present invention, the system is configured to utilize historical speech noise processing big data to collect environmental background noise feature information to achieve environmental background noise screening of the processed speech information, and accurately and reasonably extract the target speech information as a whole, efficiently and accurately achieving more effective noise reduction processing, which is an important material basis for achieving noise reduction processing of the processed speech.

[0027] The beneficial effects of the noise processing method and system based on artificial intelligence provided by the present invention are as follows:

[0028] The method extracts feature information of background noise in different environments from historical speech noise processing big data, and realizes the use of big data feature data to screen out the background noise of the processed speech. Since the extracted feature data comes from big data, the noise information loss existing in a single denoising feature can be avoided to a certain extent, so that the noise feature data in different environments can basically cover most of the noise in the corresponding environmental scenes. At the same time, on the basis of removing most of the environmental noise, the smoothing processing and semantic analysis of the speech information are used to realize the deep extraction of different types of target speech information, and it is completely extracted from a certain small amount of noise information, so as to ensure the accuracy of the extracted target speech information, make the formed speech information clearer, and effectively avoid the influence of noise on its information transmission.

[0029] The system is configured to utilize historical speech noise processing big data to collect environmental background noise feature information to achieve the screening of environmental background noise for the processed speech information, and accurately and reasonably extract the target speech information as a whole, efficiently and accurately achieving more effective noise reduction processing, which is an important material basis for achieving noise reduction processing for the processed speech. BRIEF DESCRIPTION OF THE DRAWINGS

[0030] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings required for use in the embodiments of the present invention are briefly introduced below. It should be understood that the following drawings only show certain embodiments of the present invention and therefore should not be regarded as limiting the scope. For ordinary technicians in this field, other related drawings can be obtained based on these drawings without creative work.

[0031] Figure 1 A step diagram of a noise processing method based on artificial intelligence provided by an embodiment of the present invention;

[0032] Figure 2 A schematic diagram of volume range feature extraction based on artificial intelligence provided by an embodiment of the present invention;

[0033] Figure 3 A schematic diagram of a sound wave with smoothing and denoising based on artificial intelligence provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0034] The technical solutions in the embodiments of the present invention will be described below in conjunction with the accompanying drawings in the embodiments of the present invention.

[0035] Voice collection is to extract the target voice from the complex background sound. Due to the presence of background sound, there will be a lot of noise mixed in, making the original recorded voice more mixed, and it is not possible to clearly obtain the information of the target voice. Therefore, the original recorded voice needs to be processed by noise reduction to clearly highlight the target voice.

[0036] Currently, most speech noise reduction processing is performed directly based on the large sound feature gap between the ambient background sound and the target speech. However, when this gap is relatively small, the noise reduction effect is no longer so effective, so that the noise after noise reduction still exists in the speech that affects the acquisition of the target speech, and no obvious noise reduction effect can be achieved.

[0037] refer to Figure 1 to Figure 3The embodiment of the present invention provides a noise processing method based on artificial intelligence. The method extracts feature information of background noise in different environments from historical speech noise processing big data, and uses big data feature data to screen out background noise of the processed speech. Since the extracted feature data comes from big data, the noise information loss existing in a single denoising feature can be avoided to a certain extent, so that the noise feature data in different environments can basically cover most of the noise in the corresponding environmental scenes. At the same time, on the basis of removing most of the environmental noise, the smoothing processing and semantic analysis of the speech information are used to realize the deep extraction of different types of target speech information, and it is completely extracted from a certain small amount of noise information, so as to ensure the accuracy of the extracted target speech information, make the formed speech information clearer, and effectively avoid the influence of noise on its information transmission.

[0038] The noise processing method based on artificial intelligence specifically includes the following steps:

[0039] S1: Obtain historical speech noise processing data, extract noise features based on environmental background, and form background noise feature data.

[0040] Historical speech noise processing data is obtained, and noise features are extracted based on the environmental background to form background noise feature data, including: performing volume analysis on different historical speech processing information in the historical speech noise processing data for historical speech target information to form noise volume differentiation feature data; performing background noise feature extraction on different historical speech processing information under different environmental conditions within the volume range of the historical speech target information to form target volume range background noise feature data.

[0041] The extraction of environmental background noise feature information from historical speech noise processing data mainly considers two aspects. The first aspect is that the environmental noise contained in the voice information to be processed itself, which is initially targeted in the process of historical noise removal, will have a relatively large volume of noise in the characteristic information of the sound, and there is a large gap between the sound characteristics of the target voice information to be extracted. Therefore, it can be screened in sequence. Since this obvious characteristic information gap is often contained in the collected voice data to be processed, the historical noise feature data can be used to fully extract this characteristic information, so as to quickly achieve the removal of noise with a large gap between the characteristic information and the characteristic information of the target voice information. Here, the characteristic information of the noise with a large gap is mainly considered in the volume of the noise. It can be understood that the target voice information to be collected must be clearly and completely collected during the collection process, and the clear and complete collection requires a certain range of volume. Therefore, by analyzing the volume, the noise covered by the volume that does not correspond to the volume range of the target voice information can be screened out. The second aspect is that the noise forms in different background environments are different. Not only are there great differences in non-steady-state noise, but there are also great differences in steady-state noise. Therefore, the extraction of noise feature information based on different environmental backgrounds can ensure that the extracted noise feature data is more targeted, so that the subsequent noise reduction processing can more fully achieve noise screening after reference.

[0042] The volume analysis of different historical speech processing information in the historical speech noise processing data is performed for historical speech target information to form noise volume distinguishing feature data, including: determining the volume range of historical speech target information corresponding to different historical speech processing information in the historical speech noise processing data to form the historical target speech volume range corresponding to different historical speech processing information; and performing a AND operation on all historical target speech volume ranges to form a noise volume distinguishing feature range.

[0043] Regarding the collection of noise feature information that has obvious sound feature information differences with the target voice information, we try to cover these noise feature data as much as possible under big data, so as to effectively remove this part of the noise in the subsequent process. After obtaining the noise feature information with obvious feature differences in the corresponding voice information to be processed based on the volume range of the historical voice target information, we perform a combined operation on all the obtained noise feature information to ensure that all this type of noise feature information in the historical data is covered.

[0044] Background noise feature extraction is performed on different historical voice processing information under different environmental conditions within the volume range of historical voice target information to form target volume range background noise feature data, including: extracting corresponding voice information within the volume range of historical voice target information for different historical voice processing information to form corresponding volume range historical voice information; clustering historical voice information with different volume ranges based on different background environment types to form an environment type volume range historical voice information set; performing a union operation on sound wave information of all volume range historical voice information in the volume range historical voice information set of different environment types to form target volume range background noise feature data for the corresponding background environment.

[0045] For noise with the same volume range as the target voice information, the remaining noise information is obtained by using the historical voice target information that has been calibrated in the historical voice processing information. Since there is a large gap in the noise feature information generated by different environmental backgrounds, the noise information obtained is clustered only under the same background environment. The definition of the background environment can be based on actual needs. The clustering of noise feature information based on different environmental backgrounds can further improve the pertinence of the noise feature information and improve the accuracy of the subsequent noise reduction processing of the voice information to be processed. It should be noted that for this type of noise feature information, sound wave information is used as the aspect of acquiring feature information, which can ensure that the acquired feature information is multi-dimensional, such as frequency, periodic interval, amplitude, etc. that can be obtained from the sound wave information. The comprehensive feature information is more recognizable, making the subsequent noise reduction using feature information more accurate.

[0046] S2: De-noising the speech information to be processed according to the background noise feature data to form initial de-noised speech information.

[0047] According to the background noise characteristic data, the voice information to be processed is denoised to form initial denoised voice information, including: according to the noise volume distinction characteristic range, the voice information within the noise volume distinction characteristic range is extracted to form the target processing interval voice information; according to different target volume interval background noise characteristic data, the voice information of the target processing interval is denoised with background noise to form initial denoised voice information.

[0048] After the noise feature information extraction based on the historical speech processing information is completed, the extracted noise feature information can be used to perform noise reduction on the speech information to be processed. According to the form of the extracted background noise feature data, the noise reduction of the speech information to be processed includes two sequential steps. First, the noise volume is used to distinguish the feature range to perform noise reduction based on the volume feature range on the speech information to be processed, that is, to remove the noise part with a large difference between the feature information and the feature information of the target language information. Secondly, the target volume interval background noise feature data is used to remove most of the environmental background noise of the speech information in the target processing interval. Of course, before the removal, it is necessary to determine the environmental background of the speech information in the target processing interval, and then extract the corresponding target volume interval background noise feature data for denoising.

[0049] According to different target volume interval background noise characteristic data, background noise denoising is performed on the target processing interval voice information to form initial denoised voice information, including: extracting the sound wave information of the target processing interval voice information, and matching and analyzing it with different target volume interval background noise characteristic data, determining the target volume interval background noise characteristic data with the largest number of matched sound waves, and using the background environment corresponding to the target volume interval background noise characteristic data as the background environment corresponding to the target processing interval voice information; removing the voice information corresponding to all sound wave information matching the target volume interval background noise characteristic data in the target processing interval voice information to form initial denoised voice information.

[0050] The background noise characteristic data of the corresponding target volume interval is used to denoise the background noise of the target processing interval voice information. The specific processing method is to determine whether there is noise characteristic information in the target processing interval voice information that matches the background noise characteristic data of the target volume interval. Only when there is a match can it be determined that the matched voice information belongs to environmental noise. This also avoids the loss of target voice information due to the removal of target voice information to a certain extent.

[0051] S3: Perform smoothing and semantic-based noise screening analysis on the initial denoised speech information to extract the target speech information.

[0052] The initial denoised speech information is subjected to a smoothing and semantic-based noise screening analysis to extract the target speech information, including: performing smoothing noise screening on the initial denoised speech information to determine the smoothed speech information to be confirmed, and performing a judgment analysis on the smoothed speech information to be confirmed based on music characteristics to form smooth speech type calibration information; performing semantic noise screening analysis on the initial denoised speech information to form the target language speech information.

[0053] After denoising the speech information to be processed by using the background noise feature data, the remaining speech information may still contain some new noise information such as noise that has not been covered by big data and noise whose feature information is close to the target speech feature information. Therefore, it is necessary to perform reasonable screening to extract the target speech information. Here, the target speech information is extracted from two aspects. On the one hand, if the target speech information is a speech with obvious regularity, such as different types of music, then since this regular speech is usually smoothed, the smoothness is higher than that of noise, so the analysis based on smoothness can extract speech with obvious regularity. On the other hand, if the target speech is language information, then the target speech has obvious semantic information compared to noise, and at the same time, more language information can be obtained compared to the human voice mixed in the noise. Therefore, the extraction of this type of target speech information can be achieved by analyzing the speech information.

[0054] The initial denoised speech information is subjected to smooth noise screening to determine the smooth speech information to be confirmed, and the smooth speech information to be confirmed is subjected to judgment analysis based on music characteristics to form smooth speech type calibration information, including: extracting the spectrum of all speech in the initial denoised speech information to form the original speech spectrum of the corresponding speech; derivativeizing the spectrum lines in the original speech spectrum to form the corresponding spectrum line derivative information; determining the speech information corresponding to the spectrum line with the least number of non-differentiable position points in the derivative information of different spectrum lines as the smooth speech information to be confirmed; setting the frequency change threshold, and performing the following analysis on the smooth speech information to be confirmed: if the maximum frequency difference of the spectrum line corresponding to the smooth speech information to be confirmed does not exceed the frequency change threshold, the smooth speech information to be confirmed is calibrated as non-music smooth speech information; if the maximum frequency difference of the spectrum line corresponding to the smooth speech information to be confirmed exceeds the frequency change threshold, the smooth speech information to be confirmed is calibrated as music smooth speech information.

[0055] The extraction of target speech information with regularity is mainly achieved by collecting the spectrum of all speech in the initial denoised speech information for smoothness analysis. If it is a target speech with regularity, due to its high smoothness, the derivative information formed after the spectrum derivative basically does not have many non-differentiable position points. Therefore, the speech information with the least number of statistical numbers obtained by counting the number of non-differentiable position points is most likely to be the target speech information with regularity. Of course, to determine whether there is such an obvious regularity, it is also necessary to judge the range of frequency changes. After all, there is a clear difference in the range of frequency between single tone and concert.

[0056] The initial denoised speech information is subjected to semantic noise screening analysis to form target language speech information, including: extracting language phrases from different speech information in the initial denoised speech information to form a corresponding speech phrase set; and determining the speech information with the largest number of language phrases in all speech phrase sets as the target language speech information.

[0057] The noise screening of semantic analysis of the initial denoised speech information is mainly to extract the language phrases conveyed by different speech information. Considering that if the target speech is a human voice, the phrases are coherent and obvious, the target speech information can be determined by the size of the phrases.

[0058] The present invention also provides a noise processing system based on artificial intelligence, which is configured to: obtain historical speech noise processing data, perform noise feature extraction based on environmental background, and form background noise feature data; perform denoising processing on speech information to be processed based on the background noise feature data to form initial denoised speech information; perform noise screening analysis based on smoothing and semantics on the initial denoised speech information to extract target speech information.

[0059] The system is configured to utilize historical speech noise processing big data to collect environmental background noise feature information to achieve the screening of environmental background noise for the processed speech information, and accurately and reasonably extract the target speech information as a whole, efficiently and accurately achieving more effective noise reduction processing, which is an important material basis for achieving noise reduction processing for the processed speech.

[0060] In summary, the noise processing method and system based on artificial intelligence provided by the embodiments of the present invention have the following beneficial effects:

[0061] The method extracts feature information of background noise in different environments from historical speech noise processing big data, and realizes the use of big data feature data to screen out the background noise of the processed speech. Since the extracted feature data comes from big data, the noise information loss existing in a single denoising feature can be avoided to a certain extent, so that the noise feature data in different environments can basically cover most of the noise in the corresponding environmental scenes. At the same time, on the basis of removing most of the environmental noise, the smoothing processing and semantic analysis of the speech information are used to realize the deep extraction of different types of target speech information, and it is completely extracted from a certain small amount of noise information, so as to ensure the accuracy of the extracted target speech information, make the formed speech information clearer, and effectively avoid the influence of noise on its information transmission.

[0062] The system is configured to utilize historical speech noise processing big data to collect environmental background noise feature information to achieve the screening of environmental background noise for the processed speech information, and accurately and reasonably extract the target speech information as a whole, efficiently and accurately achieving more effective noise reduction processing, which is an important material basis for achieving noise reduction processing for the processed speech.

[0063] In the embodiments of the present application, "indication" may include direct indication and indirect indication, and may also include explicit indication and implicit indication. The information indicated by a certain information is called information to be indicated. In the specific implementation process, there are many ways to indicate the information to be indicated, such as but not limited to, the information to be indicated can be directly indicated, such as the information to be indicated itself or the index of the information to be indicated. The information to be indicated can also be indirectly indicated by indicating other information, wherein there is an association between the other information and the information to be indicated. It is also possible to indicate only a part of the information to be indicated, while the other parts of the information to be indicated are known or agreed in advance. For example, the indication of specific information can also be achieved by means of the arrangement order of each piece of information agreed in advance (for example, specified by the protocol), thereby reducing the indication overhead to a certain extent. At the same time, the common parts of each piece of information can also be identified and indicated uniformly to reduce the indication overhead caused by indicating the same information separately.

[0064] In addition, the specific indication method may also be various existing indication methods, such as but not limited to the above-mentioned indication methods and various combinations thereof. The specific details of the various indication methods can refer to the prior art and will not be repeated herein. As can be seen from the above, for example, when it is necessary to indicate multiple information of the same type, different indication methods may be used for different information. In the specific implementation process, the desired indication method can be selected according to specific needs. The embodiment of the present application does not limit the selected indication method. In this way, the indication method involved in the embodiment of the present application should be understood to cover various methods that can enable the party to be indicated to obtain the information to be indicated.

[0065] It should be understood that the information to be indicated can be sent as a whole, or divided into multiple sub-information and sent separately, and the sending period and / or sending time of these sub-information can be the same or different. The specific sending method is not limited in the embodiment of the present application. Among them, the sending period and / or sending time of these sub-information can be pre-defined, for example, pre-defined according to a protocol, or can be configured by the sending end device by sending configuration information to the receiving end device.

[0066] "Pre-definition" or "pre-configuration" can be implemented by pre-saving corresponding codes, tables or other methods that can be used to indicate relevant information in the device, and the embodiments of the present application do not limit the specific implementation method. Among them, "saving" can mean saving in one or more memories. The one or more memories can be set separately or integrated in an encoder or decoder, a processor, or a communication device. The one or more memories can also be partially set separately and partially integrated in a decoder, a processor, or a communication device. The type of memory can be any form of storage medium, which is not limited by the embodiments of the present application.

[0067] The "protocol" involved in the embodiments of the present application may refer to a protocol family in the communication field, a standard protocol with a similar protocol family frame structure, or a related protocol used in future communication systems, and the embodiments of the present application do not make specific limitations on this.

[0068] In the embodiments of the present application, descriptions such as "when...", "in the case of...", "if" and "if" all mean that the device will make corresponding processing under certain objective circumstances. It does not limit the time, nor does it require the device to have a judgment action when implementing it, nor does it mean that there are other limitations.

[0069] In the description of the embodiments of the present application, unless otherwise specified, " / " indicates that the objects associated before and after are in an "or" relationship, for example, A / B can represent A or B; "and / or" in the embodiments of the present application is only a description of the association relationship of the associated objects, indicating that there can be three relationships, for example, A and / or B can represent: A exists alone, A and B exist at the same time, and B exists alone, where A and B can be singular or plural. In addition, in the description of the embodiments of the present application, unless otherwise specified, "multiple" refers to two or more than two. "At least one of the following" or its similar expression refers to any combination of these items, including any combination of single items or plural items. For example, at least one of a, b or c can represent: a, b, c, ab, ac, bc, or abc, where a, b, c can be single or multiple. In addition, in order to facilitate the clear description of the technical solution of the embodiments of the present application, in the embodiments of the present application, the words "first" and "second" are used to distinguish the same or similar items with basically the same functions and effects. Those skilled in the art will appreciate that words such as "first" and "second" do not limit the quantity and execution order, and words such as "first" and "second" do not necessarily limit the difference. At the same time, in the embodiments of the present application, words such as "exemplary" or "for example" are used to indicate examples, illustrations or explanations. Any embodiment or design described as "exemplary" or "for example" in the embodiments of the present application should not be interpreted as being more preferred or more advantageous than other embodiments or designs. Specifically, the use of words such as "exemplary" or "for example" is intended to present related concepts in a concrete way for easy understanding.

[0070] It should be understood that the processor in the embodiments of the present application may be a central processing unit (CPU), and the processor may also be other general-purpose processors, digital signal processors (DSP), application specific integrated circuits (ASIC), field programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor may be a microprocessor or the processor may also be any conventional processor, etc.

[0071] It should also be understood that the memory in the embodiments of the present application may be a volatile memory or a non-volatile memory, or may include both volatile and non-volatile memories. Among them, the non-volatile memory may be a read-only memory (ROM), a programmable read-only memory (PROM), an erasable programmable read-only memory (EPROM), an electrically erasable programmable read-only memory (EEPROM), or a flash memory. The volatile memory may be a random access memory (RAM), which is used as an external cache. By way of example and not limitation, many forms of random access memory (RAM) are available, such as static RAM (SRAM), dynamic random access memory (DRAM), synchronous DRAM (SDRAM), double data rate synchronous dynamic random access memory (DDR SDRAM), enhanced synchronous dynamic random access memory (ESDRAM), synchronous link DRAM (SLDRAM), and direct rambus RAM (DR RAM).

[0072] The above embodiments can be implemented in whole or in part by software, hardware (such as circuits), firmware or any other combination. When implemented using software, the above embodiments can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, the process or function described in the embodiment of the present application is generated in whole or in part. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium, or transmitted from one computer-readable storage medium to another computer-readable storage medium. For example, the computer instructions can be transmitted from one website site, computer, server or data center to another website site, computer, server or data center by wired (such as infrared, wireless, microwave, etc.). The computer-readable storage medium can be any available medium that a computer can access or a data storage device such as a server or data center that contains one or more available media sets. The available medium can be a magnetic medium (for example, a floppy disk, a hard disk, a tape), an optical medium (for example, a DVD), or a semiconductor medium. The semiconductor medium can be a solid-state hard disk.

[0073] It should be understood that the term "and / or" in this article is only a description of the association relationship of associated objects, indicating that there can be three relationships. For example, A and / or B can represent: A exists alone, A and B exist at the same time, and B exists alone. A and B can be singular or plural. In addition, the character " / " in this article generally indicates that the associated objects before and after are in an "or" relationship, but it may also indicate an "and / or" relationship. Please refer to the context for specific understanding.

[0074] In this application, "at least one" means one or more, and "plurality" means two or more. "At least one of the following" or similar expressions refers to any combination of these items, including any combination of single or plural items. For example, at least one of a, b, or c can mean: a, b, c, ab, ac, bc, or abc, where a, b, c can be single or multiple.

[0075] It should be understood that in the various embodiments of the present application, the size of the serial numbers of the above-mentioned processes does not mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present application.

[0076] Those of ordinary skill in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of this application.

[0077] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices and units described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0078] In the several embodiments provided in the present application, it should be understood that the disclosed systems, devices and methods can be implemented in other ways. For example, the device embodiments described above are only schematic. For example, the division of the units is only a logical function division. There may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.

[0079] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0080] In addition, each functional unit in each embodiment of the present application may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit.

[0081] If the functions are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application can be essentially or partly embodied in the form of a software product that contributes to the prior art. The computer software product is stored in a storage medium and includes several instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the methods described in the various embodiments of the present application. The aforementioned storage media include: various media that can store program codes, such as USB flash drives, mobile hard disks, read-only memories (ROM), random access memories (RAM), magnetic disks or optical disks.

[0082] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art who is familiar with the present technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

Claims

1. A noise processing method based on artificial intelligence, characterized in that: include: Acquire historical speech noise processing data, extract noise features based on environmental background, and form background noise feature data; According to the background noise characteristic data, the voice information to be processed is subjected to denoising processing to form initial denoised voice information; The initial denoised speech information is subjected to noise screening analysis based on smoothing and semantics to extract target speech information.

2. The noise processing method based on artificial intelligence according to claim 1, characterized in that: The step of obtaining historical speech noise processing data and extracting noise features based on environmental background to form background noise feature data includes: Performing volume analysis on different historical speech processing information in the historical speech noise processing data with respect to historical speech target information to form noise volume distinguishing feature data; Background noise feature extraction is performed on different historical speech processing information under different environmental conditions within the volume range of the historical speech target information to form target volume interval background noise feature data.

3. The noise processing method based on artificial intelligence according to claim 2, characterized in that: The performing volume analysis on different historical speech processing information in the historical speech noise processing data for historical speech target information to form noise volume distinguishing feature data includes: Determining the volume range of the historical speech target information corresponding to different historical speech processing information in the historical speech noise processing data, and forming the historical target speech volume range corresponding to different historical speech processing information; All the historical target speech volume ranges are combined to form a noise volume distinguishing feature range.

4. The noise processing method based on artificial intelligence according to claim 3 is characterized in that: The extracting of background noise features under different environmental conditions within the volume range of the historical speech target information from different historical speech processing information to form background noise feature data in the target volume range includes: For different historical voice processing information, extracting corresponding voice information within the volume range of the historical voice target information to form corresponding volume range historical voice information; Clustering the historical voice information of different volume ranges based on different background environment types to form a historical voice information set of environment type volume range; A sound wave information union operation is performed on all the volume range historical voice information in the volume range historical voice information set of different environment types to form the target volume interval background noise characteristic data of the corresponding background environment.

5. The noise processing method based on artificial intelligence according to claim 4 is characterized in that: The method of performing denoising on the speech information to be processed according to the background noise feature data to form initial denoised speech information includes: According to the noise volume distinguishing feature range, extracting the voice information within the noise volume distinguishing feature range from the voice information to be processed to form the target processing interval voice information; According to the different background noise characteristic data of the target volume interval, the background noise is denoised for the speech information in the target processing interval to form initial denoised speech information.

6. The noise processing method based on artificial intelligence according to claim 5, characterized in that: The method of performing background noise denoising on the speech information in the target processing interval according to the background noise characteristic data of the different target volume intervals to form initial denoised speech information includes: Extracting the sound wave information of the target processing interval voice information, and matching and analyzing it with different target volume interval background noise feature data, determining the target volume interval background noise feature data with the largest number of matching sound waves, and using the background environment corresponding to the target volume interval background noise feature data as the background environment corresponding to the target processing interval voice information; The speech information corresponding to all the sound wave information matching the background noise feature data in the target volume interval is removed from the speech information in the target processing interval to form the initial denoised speech information.

7. The noise processing method based on artificial intelligence according to claim 6, characterized in that: The performing of smoothing and semantic-based noise screening analysis on the initial denoised speech information to extract target speech information includes: Performing smooth noise screening on the initial denoised voice information to determine smooth voice information to be confirmed, and performing judgment analysis on the smooth voice information to be confirmed based on music characteristics to form smooth voice type calibration information; The initial denoised speech information is subjected to semantic noise screening analysis to form target language speech information.

8. The noise processing method based on artificial intelligence according to claim 7, characterized in that: The smoothing noise screening of the initial denoised voice information to determine the smoothed voice information to be confirmed, and the judgment analysis of the smoothed voice information to be confirmed based on the music characteristics to form the smoothed voice type calibration information, including: Extracting the frequency spectra of all the speech in the initial denoised speech information to form the original speech spectrum of the corresponding speech; Derivativeizing the spectrum lines in the original speech spectrum to form corresponding spectrum line derivative information; Determine the speech information corresponding to the spectrum line having the least number of position points that cannot be derived in the different spectrum line derivative information as the smoothed speech information to be confirmed; The frequency change threshold is set, and the smoothed voice information to be confirmed is analyzed as follows: If the maximum frequency difference of the frequency spectrum line corresponding to the smoothed voice information to be confirmed does not exceed the frequency change threshold, the smoothed voice information to be confirmed is marked as non-music smoothed voice information; If the maximum frequency difference of the frequency spectrum lines corresponding to the smoothed voice information to be confirmed exceeds the frequency change threshold, the smoothed voice information to be confirmed is marked as music smoothed voice information.

9. The noise processing method based on artificial intelligence according to claim 8, characterized in that: The performing semantic noise screening analysis on the initial denoised speech information to form target language speech information includes: Extracting language phrases from different voice information in the initial denoised voice information to form a corresponding voice phrase set; The speech information with the largest number of language phrases in all the speech phrase sets is determined as the target language speech information.

10. A noise processing system based on artificial intelligence, characterized in that: Configured as: Acquire historical speech noise processing data, extract noise features based on environmental background, and form background noise feature data; According to the background noise characteristic data, the voice information to be processed is subjected to denoising processing to form initial denoised voice information; The initial denoised speech information is subjected to noise screening analysis based on smoothing and semantics to extract target speech information.

Images