Sound masking processing method, device, carrier and storage medium

Abstract

This application relates to a sound masking processing method, apparatus, carrier, and storage medium. The method includes: acquiring a target speech segment; performing sound masking processing on at least one speech feature of the target speech segment to obtain a masking signal; and outputting the masking signal. The sound masking processing method provided by this application, by performing sound masking processing on at least one speech feature of the target speech segment, enables the obtained masking signal to meet low latency requirements and improve user experience.

Description

Technical Field

[0001] The present application relates to the field of speech signal processing, and in particular to a sound masking processing method, device, carrier and storage medium. Background Art

[0002] The existing method of masking sound to prevent information leakage mainly involves time-flipping the acquired voice clips and directly outputting the flipped voice clips as masking signals. However, this method requires the frame length of the acquired voice clips to be in the range of 100ms to 500ms, which results in high latency in the existing sound masking methods, resulting in a poor user experience. Summary of the invention

[0003] The embodiments of the present application provide a sound masking processing method, device, carrier and storage medium, which perform sound masking processing on at least one speech feature of a target speech segment so that the obtained masking signal meets the low delay requirement and improves the user experience.

[0004] In order to achieve the above object, according to a first aspect of the present application, a sound masking processing method is provided, the method comprising:

[0005] Get the target speech segment;

[0006] Performing sound masking processing on at least one speech feature of the target speech segment to obtain a masking signal;

[0007] The mask signal is output.

[0008] In some embodiments, the speech feature includes a power spectrum, and the performing sound masking processing on at least one speech feature of the target speech segment to obtain a masking signal includes: performing power spectrum analysis on the target speech segment to obtain a power spectrum; performing spectrum cropping processing based on the power spectrum to obtain a cropped power spectrum; and obtaining a masking signal based on the cropped power spectrum.

[0009] In some embodiments, the spectral clipping processing is performed based on the power spectrum to obtain a clipped power spectrum, including: dividing the power spectrum into segments according to the amplitude values ​​in the power spectrum to obtain at least one power spectrum sub-segment, each power spectrum sub-segment corresponding to an amplitude range; in the power spectrum, clipping the power spectrum sub-segment whose amplitude range is smaller than a preset amplitude value to obtain a clipped power spectrum.

[0010] In some embodiments, the speech feature includes phase information, and performing sound masking processing on at least one speech feature of the target speech segment to obtain a masking signal includes: acquiring the phase information of the target speech segment; and adding phase noise to the target speech segment based on the phase information to obtain a masking signal.

[0011] In some embodiments, before performing sound masking processing on at least one speech feature of the target speech segment to obtain a masking signal, it also includes: performing time flipping processing on the target speech segment to obtain a flipped speech segment; performing sound masking processing on at least one speech feature of the target speech segment to obtain a masking signal includes: performing sound masking processing on at least one speech feature of the flipped speech segment to obtain a masking signal.

[0012] In some embodiments, before performing sound masking processing on at least one speech feature of the target speech segment to obtain a masking signal, it also includes: performing feature filtering processing on the target speech segment to obtain a filtered speech segment; performing sound masking processing on at least one speech feature of the target speech segment to obtain a masking signal includes: performing sound masking processing on at least one speech feature of the filtered speech segment to obtain a masking signal.

[0013] In some embodiments, the feature filtering processing of the target speech segment to obtain a filtered speech segment includes: performing amplitude spectrum analysis on the target speech segment to obtain an amplitude spectrum; performing feature extraction on the target speech segment and generating a feature filter based on the extracted features; and performing feature filtering processing on the target speech segment based on the feature filter and the amplitude spectrum to obtain a filtered speech segment.

[0014] In some embodiments, the feature extraction of the target speech segment and the generation of a feature filter based on the extracted features include: performing a cepstrum analysis on the target speech segment to obtain a fundamental frequency and at least one formant feature; and generating a feature filter based on the fundamental frequency and formant features.

[0015] In some embodiments, obtaining the masking signal based on the cropped power spectrum includes: obtaining a cropped processed speech segment based on the cropped power spectrum; adding noise to the cropped processed speech segment to obtain a processed speech segment with noise added; and obtaining the masking signal based on the processed speech segment with noise added.

[0016] In some embodiments, the added noise includes at least one of frequency domain noise and time domain noise;

[0017] The step of performing noise adding processing on the processed voice segment after trimming to obtain the processed voice segment after noise addition comprises: when the added noise is frequency domain noise, performing spectrum analysis on the processed voice segment after trimming to obtain a spectrum, and based on the spectrum and the frequency domain noise, performing frequency domain noise adding processing on each voice data in the processed voice segment after trimming to obtain the processed voice segment after noise addition; when the added noise is time domain noise, performing time domain noise adding processing on each voice data in the processed voice segment after trimming in the time domain based on the time domain noise to obtain the processed voice segment after noise addition.

[0018] In some embodiments, obtaining a masking signal based on the processed speech segment after adding noise includes: obtaining a preset target masking ratio; determining a gain factor of each speech data in the processed speech segment after adding noise based on the preset target masking ratio; performing gain control on the processed speech segment after adding noise based on the gain factor of each speech data to obtain a gain speech segment; and obtaining a masking signal based on the gain speech segment.

[0019] In some embodiments, the obtaining of the target voice segment includes: obtaining, by a voice collection module disposed in a first space of the vehicle, a voice emitted by a user in the first space to obtain the target voice segment;

[0020] The outputting the masking signal includes: outputting the masking signal to the second space through a voice playing module arranged in the second space of the vehicle.

[0021] According to a second aspect of the present application, a sound masking processing device is provided, the device comprising:

[0022] A voice collection module is used to obtain target voice segments;

[0023] A masking signal generating module, configured to perform sound masking processing on at least one speech feature of the target speech segment to obtain a masking signal;

[0024] The voice playing module is used to output the masking signal.

[0025] According to a third aspect of the present application, a vehicle is provided, comprising a first space and a second space, wherein the first space is provided with the above-mentioned voice collection module, the second space is provided with the above-mentioned voice playback module, and the above-mentioned masking signal generation module is also provided on the vehicle.

[0026] According to a fourth aspect of the present application, a computer-readable storage medium is provided, wherein the computer-readable storage medium stores a computer program. When the computer program is run on a vehicle, the computer program is used to enable the vehicle to execute any one of the sound masking processing methods provided in the embodiments of the present application.

[0027] According to a fifth aspect of the present application, a computer program product is provided, including a computer program, and when the computer program is executed by a processor, any one of the sound masking processing methods provided in the embodiments of the present application is implemented.

[0028] In the sound masking processing method, device, carrier and storage medium of the embodiment of the present application, a target voice segment is obtained; at least one voice feature of the target voice segment is subjected to sound masking processing to obtain a masking signal; and the masking signal is output. By performing sound masking processing on at least one voice feature of the target voice segment, the obtained masking signal meets the low latency requirement and improves the user experience. BRIEF DESCRIPTION OF THE DRAWINGS

[0029] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For those skilled in the art, other drawings can be obtained based on these drawings without creative work.

[0030] Figure 1 It is a flowchart of the sound masking processing method provided by the present application;

[0031] Figure 2 It is a structural schematic diagram of a vehicle carrier provided by the present application;

[0032] Figure 3 It is a flowchart of a specific embodiment of a sound masking processing method provided in an embodiment of the present application;

[0033] Figure 4 It is a flow chart of a characteristic filtering process of a sound masking processing method provided by an embodiment of the present application;

[0034] Figure 5 It is a flow chart of a spectrum clipping process of a sound masking processing method provided by an embodiment of the present application;

[0035] Figure 6 The embodiment of the present application provides a performance diagram of different sound masking schemes under a fixed TMR condition;

[0036] Figure 7 is a structural schematic diagram of a sound masking processing device provided in an embodiment of the present application;

[0037] Figure 8 It is a structural schematic diagram of a masking signal generating module provided in an embodiment of the present application. DETAILED DESCRIPTION

[0038] The technical solutions in the embodiments of the present application will be described clearly and completely below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative work are within the scope of protection of the present application.

[0039] In the description of this application, the terms "first" and "second" are used for descriptive purposes only and should not be understood as indicating or implying relative importance or implicitly indicating the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more features. In the description of this application, the meaning of "plurality" is two or more, unless otherwise clearly and specifically defined.

[0040] See also Figure 1 , Figure 1 : is a flow chart of a sound masking processing method provided by an embodiment of the present application. The sound masking processing method may include the following steps S101 to S103:

[0041] Step S101, obtaining a target voice segment.

[0042] The target speech segment refers to the speech segment that needs to be subjected to sound masking processing.

[0043] There are many ways to obtain the target voice segment, which are not limited in the embodiments of the present application. For example, the voice to be masked is collected in real time and transmitted to the masking signal generation module in the form of digital signals in frames, and the target voice segment is determined from the voice to be masked in the masking signal generation module according to a preset frame length (such as 40ms). For another example, the voice acquisition module set in the first space of the vehicle is used to obtain the voice uttered by the user in the first space to obtain the target voice segment.

[0044] Step S102: Perform sound masking processing on at least one speech feature of the target speech segment to obtain a masking signal.

[0045] Among them, sound masking processing refers to a series of processing performed to mask specific voice information.

[0046] Among them, the speech feature refers to the feature of the sound masking process for the target speech segment. There are many options for speech features, which can be adjusted according to actual conditions, and the embodiments of the present application do not limit it. For example, the speech feature is the power spectrum. For another example, the speech feature is the phase information.

[0047] By performing sound masking processing on at least one voice feature of the target voice segment, a suitable masking signal is generated. Through masking signal interference, other sounds are applied to the location where there is a risk of eavesdropping to interfere with the receiver's understanding of the leaked voice. Through sound masking processing, there is no need to perform complex acoustic modifications on the vehicle. Instead, by designing appropriate sound masking processing, the intelligibility of the leaked voice to potential eavesdroppers can be effectively reduced under the action of psychoacoustic masking effect, thereby achieving masking and effectively preventing message leakage.

[0048] Step S103: outputting a masking signal.

[0049] In the sound masking processing method of the embodiment of the present application, a target speech segment is obtained; at least one speech feature of the target speech segment is subjected to sound masking processing to obtain a masking signal; and the masking signal is output. By performing sound masking processing on at least one speech feature of the target speech segment, the obtained masking signal meets the low latency requirement and improves the user experience.

[0050] In some embodiments, a voice acquisition module provided in the first space of the vehicle can be used to acquire the voice of the user in the first space to obtain the target voice segment, and a voice playback module provided in the second space of the vehicle can be used to output a masking signal to the second space.

[0051] The vehicle is used to indicate the vehicle used in the application scenario that needs to be sound masked. The vehicle includes at least one of a vehicle vehicle, a flight vehicle, a motor vehicle vehicle, and a ship vehicle. For example, the vehicle used in the car scenario is a vehicle. The vehicle used in the flight scenario is an airplane.

[0052] The common features of the vehicles are as follows: the vehicle comprises a first space and a second space, a voice collection module is provided in the first space, a voice playback module is provided in the second space, and the vehicle is also provided with a masking signal generation module.

[0053] The first space is used to indicate a spatial area protected by voice privacy, and the second space is used to indicate a spatial area where a masking signal is played.

[0054] like Figure 2 As shown, Figure 2 It is a structural schematic diagram of a vehicle carrier provided in an embodiment of the present application. Figure 2 The seating area in the rear cabin is the privacy protection area, which specifically refers to the seating area in the car that is protected by voice privacy, that is, the area where private voice is generated, that is, the first space. Figure 2 The front cockpit seat area is the voice isolation area, which specifically refers to the seat area in the car where private voices are not expected to be leaked, that is, the area where interference signals are masked, that is, the second space. Figure 2The first space shown is equipped with a microphone, namely, a voice collection module. Figure 2 The second space shown is equipped with a headrest speaker, i.e., a voice playback module. Figure 2 The vehicle-mounted computing platform shown is configured with a masking signal generating module.

[0055] The voice to be masked in the privacy protection zone is collected by the microphone inside the car and transmitted in frames to the masking signal generation module in the form of digital signals. The masking signal is calculated and output by the on-board computing platform, and finally played by the headrest speakers in the voice isolation zone, thereby interfering with the voice reception of passengers in the voice isolation zone.

[0056] In this embodiment, a voice acquisition module is provided in the first space of the vehicle to acquire the voice uttered by the user in the first space to obtain a target voice segment, and a voice playback module is provided in the second space of the vehicle to output a masking signal to the second space to perform sound masking processing on at least one voice feature of the target voice segment so that the obtained masking signal meets the low latency requirement and improves the user experience.

[0057] It is understandable that there are multiple options for performing sound masking processing on at least one speech feature of the target speech segment to obtain a masking signal, which can be selected according to actual conditions and are not limited in the embodiments of the present application.

[0058] In some embodiments, the speech feature includes a power spectrum, and the process of performing sound masking processing on at least one speech feature of the target speech segment to obtain a masking signal may include: performing power spectrum analysis on the target speech segment to obtain a power spectrum; performing spectrum cropping processing based on the power spectrum to obtain a cropped power spectrum; and obtaining a masking signal based on the cropped power spectrum.

[0059] The power spectrum refers to the spectrum of signal power changing with frequency. By using the fast Fourier transform, the sound signal in the target speech segment in the time domain is converted into a sound signal in the frequency domain. The target speech segment after the frequency domain transformation is subjected to spectrum analysis to obtain the spectrum, and the amplitude corresponding to the spectrum is squared to obtain the power spectrum.

[0060] The process of obtaining the cropped processed speech segment based on the cropped power spectrum is as follows: performing an inverse transformation on the cropped power spectrum to obtain the cropped processed speech segment.

[0061] In this embodiment, a power spectrum is obtained by performing power spectrum analysis on the target speech segment; spectrum cropping is performed based on the power spectrum to obtain a cropped power spectrum; and a masking signal is obtained based on the cropped power spectrum, thereby removing redundant features of the speaker and reducing the semantic intelligibility of the speech segment itself, thereby improving masking performance while meeting low latency.

[0062] Furthermore, in some optional embodiments, spectrum clipping is performed based on the power spectrum to obtain a clipped power spectrum. The process may be: according to the amplitude value in the power spectrum, the power spectrum is segmented to obtain at least one power spectrum sub-segment, and one power spectrum sub-segment corresponds to an amplitude range; in the power spectrum, the power spectrum sub-segment whose amplitude range is smaller than a preset amplitude value is clipped to obtain a clipped power spectrum.

[0063] Among them, for different speech segments, the preset amplitude values ​​can be the same or different, and can be adjusted according to actual conditions, which is not limited in the embodiments of the present application. The source of the preset amplitude value can be manually preset or calculated by a spectrum clipping function. There are many ways to set the spectrum clipping function, which can be adjusted according to actual conditions, which is not limited in the embodiments of the present application. For example, the spectrum clipping function is set to an exponential function form with 2 as the base. For another example, the spectrum clipping function is set to an exponential function form with 10 as the base.

[0064] Exemplarily, it is assumed that the power spectrum can be divided into N=10 segments according to its amplitude value. Where the maximum value of the power spectrum amplitude is Max_P, then segment 0 represents the part of the power spectrum with an amplitude value less than 0.1×Max_P, segment 1 represents the part of the power spectrum with an amplitude value less than 0.2×Max_P and greater than 0.1×Max_P, and so on. The corresponding spectrum clipping function is set to an exponential function with a base of 2. Through the spectrum clipping function, it can be determined that the power spectrum of segments n=1~10 is scaled by multiples of 0.03125, 0.0625, 0.125, 0.25, 0.5, 1, 2, 4, 8, and 16, respectively. That is, the 6th segment can be used as the preset amplitude value, the part with an amplitude value lower than 0.5×Max_P is clipped, and the part with an amplitude value higher than 0.6×Max_P is amplified (enhanced).

[0065] In this embodiment, the power spectrum is segmented according to the amplitude values ​​in the power spectrum to obtain at least one power spectrum sub-segment, and one power spectrum sub-segment corresponds to an amplitude range; in the power spectrum, the power spectrum sub-segments with amplitude ranges smaller than the preset amplitude value are clipped to obtain a clipped power spectrum, so as to analyze the power spectrum of the target speech segment, and remove redundant features of the speaker by retaining and enhancing the partial spectrum exceeding the power threshold and suppressing the partial spectrum below the power threshold, thereby reducing the semantic intelligibility of the masking signal obtained based on the clipped processed speech segment, thereby improving the masking performance while meeting the low latency requirement.

[0066] In some embodiments, the speech feature includes phase information, and the process of performing sound masking processing on at least one speech feature of the target speech segment to obtain a masking signal may also include: acquiring the phase information of the target speech segment; adding phase noise to the target speech segment based on the phase information to obtain a masking signal.

[0067] Phase information refers to the information obtained after taking the phase angle of the signal in the frequency domain. Phase noise refers to the random change of the phase of the masking signal output by the target speech segment caused by various noises.

[0068] In this embodiment, a target speech segment is obtained, and phase information of the target speech segment is obtained; phase noise is added to the target speech segment based on the phase information to obtain a masking signal. By adding phase noise, the semantic intelligibility of the speech segment itself is reduced, and the masking performance is improved while meeting the low latency requirement.

[0069] In some embodiments, the process of performing sound masking processing on at least one speech feature of the target speech segment to obtain a masking signal may also include: performing time flipping processing on the target speech segment to obtain a flipped speech segment, and performing sound masking processing on at least one speech feature of the flipped speech segment to obtain a masking signal.

[0070] Among them, time flipping refers to flipping the time domain signal in the speech segment.

[0071] It should be noted that time flipping can occur before spectral cropping based on the power spectrum of the target speech segment. Before spectral cropping based on the power spectrum of the target speech segment, it is necessary to transform the flipped speech segment to obtain a flipped speech segment in the frequency domain, and then perform spectral cropping based on the power spectrum of the flipped speech segment in the frequency domain. Time flipping can also occur after spectral cropping based on the power spectrum of the target speech segment. After spectral cropping based on the power spectrum of the target speech segment, it is necessary to perform an inverse fast Fourier transform on the frequency domain signal in the cropped processed speech segment, convert the frequency domain signal in the processed speech segment into a time domain signal, and then perform flipping on the speech segment in the time domain.

[0072] The process of time flipping is: according to the time axis where each voice data of the target voice segment is located, the voice data is reversed to rearrange the voice data to obtain the flipped voice segment. For example, the target voice segment includes 10 voice data, and their time information is 0 to 9 respectively. The time information of the 0th voice data is changed to 9, and the time information of the 1st voice data is changed to 8, and so on, and the voice data are rearranged to obtain the flipped voice segment.

[0073] Each speech segment contains a number of speech data. For example, for a 40ms speech segment, assuming the sampling rate is 16000Hz, the data points contained in this speech segment are sampling rate × frame length (0.04s) = 640, and the time reversal processing is to reverse the order of these 640 data.

[0074] In this embodiment, a target speech segment is acquired, time-flipped to the target speech segment to obtain a flipped speech segment, and sound masking is performed on at least one speech feature of the flipped speech segment to obtain a masking signal; the masking signal is output to obtain a preliminary human-like masking signal, thereby reducing the semantic intelligibility of the speech segment itself, thereby improving the masking performance while meeting the low latency requirement.

[0075] In some embodiments, before performing sound masking processing on at least one speech feature of the target speech segment to obtain a masking signal, the method further includes: performing feature filtering processing on the target speech segment to obtain a filtered speech segment. The process of performing feature filtering processing on the target speech segment to obtain a filtered speech segment may also include: performing sound masking processing on at least one speech feature of the filtered speech segment to obtain a masking signal.

[0076] Among them, feature filtering is an operation performed to enhance the correlation between the current target speech segment and the corresponding original speech segment.

[0077] In this embodiment, a target speech segment is obtained, feature filtering is performed on the target speech segment to obtain a filtered speech segment; at least one speech feature of the filtered speech segment is subjected to acoustic masking processing to obtain a masking signal; and the masking signal is output. Through the above process, the correlation between the current target speech segment and the corresponding original speech segment is enhanced, so that the masking performance is improved while meeting the low delay.

[0078] In one embodiment, the process of performing feature filtering on the target speech segment to obtain a filtered speech segment may be: performing amplitude spectrum analysis on the target speech segment to obtain an amplitude spectrum; performing feature extraction on the target speech segment and generating a feature filter based on the extracted features; and performing feature filtering on the target speech segment based on the feature filter and the amplitude spectrum to obtain a filtered speech segment.

[0079] The amplitude spectrum is used to indicate the spectrum of signal amplitude changing with frequency. The process of amplitude spectrum analysis can be: performing Fourier transform on the digital signal corresponding to the voice data, taking the absolute value of the complex spectrum signal after Fourier transform, and then obtaining the amplitude spectrum.

[0080] The feature filter is used to indicate a filter configured according to the features extracted from the target speech segment. The feature filter includes multiple coefficients, which are determined according to the features extracted from the processed speech segment, and the embodiment of the present application does not limit this.

[0081] In this embodiment, the characteristic filtering of the target speech segment can be completed by multiplying each coefficient of the characteristic filter with the corresponding item of the amplitude spectrum.

[0082] In this embodiment, the amplitude spectrum is analyzed on the target speech segment to obtain the amplitude spectrum; the feature extraction is performed on the target speech segment, and a feature filter is generated according to the extracted features; based on the feature filter and the amplitude spectrum, the target speech segment is subjected to feature filtering processing to obtain a filtered speech segment. Through the above process, the target speech segment is subjected to feature filtering processing based on the obtained feature filter and the amplitude spectrum to obtain a filtered speech segment, so as to better enhance the correlation between the current target speech segment and the corresponding original speech segment, so as to improve the masking performance while meeting the low delay.

[0083] There are many ways to generate the feature filter. For example, a human voice corpus is constructed based on the target speech segment. The average speech spectrum is calculated based on the speech data of the target speech segment and the human voice corpus. The feature filter is generated based on the feature information contained in the average speech spectrum. For another example, a cepstrum analysis is performed on the target speech segment to obtain a fundamental frequency and at least one formant feature; and a feature filter is generated based on the fundamental frequency and formant features.

[0084] In this example, cepstrum analysis refers to a spectrum analysis method in speech signal processing. The analysis process is: Fourier transform spectrum of the digital signal corresponding to the speech data is logarithmically calculated and then Fourier inverse transform is performed. The fundamental frequency is a speech signal feature that reflects the vibration characteristics of the speaker's glottis. The formant is a speech signal feature that reflects the filtering characteristics of the speaker's vocal tract.

[0085] In this example, a feature filter is generated by the speaker's fundamental frequency and formant characteristics, and the amplitude spectrum of the processed speech segment is feature filtered based on the feature filter. This can enhance the correlation between the masking signal and the speaker's glottal impulse and vocal tract response characteristics, thereby improving the masking performance while meeting low latency.

[0086] The following is a specific example to explain feature filtering. Assume that after the processed speech segment is subjected to cepstrum analysis, the speaker's fundamental frequency f0 and two formant features are extracted. The formant features include formant f1 and formant peak value A1 corresponding to formant f1, and formant f2 and formant peak value A2 corresponding to formant f2. Based on the extracted fundamental frequency f0 and two formant features, feature filtering is performed on the speech segment according to the following formula (1):

[0087] (1)

[0088] in, Used to indicate the amplitude spectrum, which is formed by the superposition of three Hamming window functions. BW0 refers to the Hamming window bandwidth of the characteristic filter for frequency point f0, BW1 refers to the Hamming window bandwidth of the characteristic filter for frequency point f1, BW2 refers to the Hamming window bandwidth of the characteristic filter for frequency point f2, and the window peak value for the base frequency f0 is A0.

[0089] In some embodiments, the process of performing acoustic masking processing on at least one speech feature of the target speech segment to obtain a masking signal may be: performing feature filtering on the target speech segment to obtain a filtered speech segment, and performing spectrum clipping processing based on the power spectrum of the filtered speech segment. The process of performing acoustic masking processing on at least one speech feature of the target speech segment to obtain a masking signal may also be: performing time flipping processing on the target speech segment to obtain a flipped speech segment, performing feature filtering on the flipped speech segment to obtain a filtered speech segment, performing spectrum clipping processing based on the power spectrum of the filtered speech segment to obtain a clipped processed segment, and obtaining a masking signal based on the clipped processed speech segment.

[0090] In some embodiments, the process of obtaining the masking signal based on the cropped power spectrum can be: obtaining a cropped processed speech segment based on the cropped power spectrum; adding noise to the cropped processed speech segment to obtain a processed speech segment with noise added; and obtaining the masking signal based on the processed speech segment with noise added.

[0091] It should be noted that there are multiple options for the timing of noise addition, which can be adjusted according to actual conditions. For example, the noise addition process is directly performed on the flipped speech segment after time flipping. For another example, the noise addition process is directly performed on the filtered speech segment after feature filtering. For another example, the noise addition process is directly performed on the processed speech segment after cropping.

[0092] Preferably, the embodiment of the present application adopts the method of directly adding noise to the processed speech segment after cropping, so as to further reduce the semantic comprehensibility of the masking signal itself obtained from the processed speech segment after noise addition processing while removing redundant features of the speaker and reducing the semantic comprehensibility of the masking signal itself obtained from the processed speech segment after spectral cropping, so as to improve the masking performance while meeting the low delay.

[0093] The added noise includes at least one of frequency domain noise and time domain noise.

[0094] Frequency domain noise is used to indicate the noise that interferes or adds to the signal in the frequency domain. It is expressed by introducing energy or frequency changes in the signal's spectrum. Frequency domain noise includes but is not limited to white noise, narrowband noise, and impulse noise.

[0095] Specifically, when the added noise is frequency domain noise, a spectrum analysis is performed on the cropped processed speech segment to obtain a spectrum, and based on the spectrum and the frequency domain noise, frequency domain noise is added to each speech data in the cropped processed speech segment to obtain a processed speech segment with noise added.

[0096] Preferably, the frequency domain noise in the embodiment of the present application adopts multiplicative Gaussian white noise.

[0097] Time domain noise refers to the noise that interferes or adds to the signal in the time domain, which is displayed by the change of the signal amplitude value or the jump at the moment. Time domain noise includes but is not limited to Gaussian noise, impulse noise, periodic noise, and continuous steady-state noise in nature.

[0098] Specifically, when the added noise is time domain noise, based on the time domain noise, each voice data in the clipped processed voice segment in the time domain is subjected to time domain noise addition processing to obtain a processed voice segment with added noise.

[0099] Preferably, the time domain noise in the embodiment of the present application is continuous steady-state noise in nature, and the weight ratio of the processed speech segment in the time domain and the continuous steady-state noise in nature when superimposed is 1:1.

[0100] By adding noise to the processed speech segment, the semantic intelligibility of the masking signal obtained from the processed speech segment is further reduced, so that the masking performance is improved while meeting the low latency requirement.

[0101] In some embodiments, frequency domain noise and time domain noise can also be added at the same time. If the order of adding noise is: frequency domain noise, time domain noise, it is necessary to convert the frequency domain signal in the processed voice segment after adding the frequency domain noise into the time domain signal, and then add the time domain noise to the processed voice segment in the time domain. If the order of adding noise is: time domain noise, frequency domain noise, it is necessary to convert the time domain signal in the processed voice segment after adding the time domain noise into the frequency domain signal, and then add the frequency domain noise to the processed voice segment in the frequency domain.

[0102] In some embodiments, gain control can also be performed on the target speech segment. There are multiple options for the timing of gain control processing, which can be adjusted according to actual conditions. For example, gain control is directly performed on the processed speech segment after time flipping. For another example, gain control is directly performed on the processed speech segment after filtering. For another example, gain control is directly performed on the processed speech segment after spectral cropping. For another example, gain control is performed on the processed speech segment after noise addition processing.

[0103] Preferably, the embodiment of the present application adopts direct gain control on the processed speech segment after the noise addition process.

[0104] Among them, the specific process of obtaining the masking signal based on the processed speech segment after noise addition is: obtaining a preset target masking ratio; based on the preset target masking ratio, determining the gain factor of each speech data in the processed speech segment after noise addition; based on the gain factor of each speech data, performing gain control on the processed speech segment after noise addition to obtain a gained speech segment; based on the gained speech segment, obtaining the masking signal.

[0105] In this embodiment, the preset target masking ratio refers to a preset target masking ratio (TMR). The preset target masking ratio can be set according to actual conditions, for example, directly setting a specific preset target masking ratio. For another example, the preset target masking ratio is set according to relevant information of the speech segment.

[0106] For example, according to the following formula (2), the preset target masking ratio is determined according to the sampling rate and frame length of the speech segment:

[0107] (2)

[0108] TMR refers to the preset target masking ratio, which is expressed in decibels (dB). T(n) refers to the original speech data of n points in length, that is, the speech data contained in the processed segment, and M(n) refers to the masked signal frame of n points in length, that is, the masked signal frame corresponding to the speech data contained in the processed segment. n represents the number of sampling points corresponding to the frame length. For example, if the sampling rate is 16000Hz and the frame length is 40ms, then n=0.04*16000=640.

[0109] In this embodiment, the gain factor is a factor for performing gain control on the corresponding voice data. The gain factor can be adjusted according to a preset target masking ratio.

[0110] For example, by presetting the target masking ratio, the actual TMR of each speech data in the processed speech segment is calculated and set to R.

[0111] Furthermore, the gain factor corresponding to each voice data is determined according to the following formula (3):

[0112] (3)

[0113] Wherein, G refers to the gain factor corresponding to the speech data, R refers to the actual TMR of the speech data, and TMR refers to the target masking ratio.

[0114] Preferably, the preset target masking signal ratio TMR is -10dB, then the masking signal gain factor is .

[0115] In this embodiment, the masking signal may be outputted by multiplying each masking signal frame by a gain factor G of the corresponding speech data.

[0116] In this embodiment, a processed speech segment is obtained by acquiring a target speech segment and performing sound masking processing on at least one speech feature of the target speech segment; a preset target masking ratio is obtained; based on the preset target masking ratio, a gain factor of each speech data in the processed speech segment after noise addition is determined; based on the gain factor of each speech data, a gain control is performed on the processed speech segment after noise addition to obtain a gained speech segment; based on the gained speech segment, a masking signal is obtained to improve the masking performance.

[0117] In some embodiments, the processed speech segment may also be subjected to windowing and smoothing processing, and the windowing and smoothing processing may be performed before the masking signal is obtained.

[0118] Specifically, the process of windowing and smoothing processing is: obtaining a preset window function, and based on the preset window function, performing windowing and smoothing processing on the processed speech segment to generate a masking signal.

[0119] There are multiple options for the preset window function, which can be adjusted according to actual conditions, and the embodiments of the present application do not limit it. For example, the preset window function is a Hanning window. For another example, the preset window function is a Hamming window. For another example, the preset window function is a Gaussian window. For another example, the preset window function is other composite windows.

[0120] The process of windowing and smoothing can be adjusted according to actual conditions. For example, the masking signal corresponding to each voice data in the processed voice segment is directly multiplied by a preset window function and then output.

[0121] In a specific embodiment, in order to solve the problems of high masking signal output delay, low masking sound speech intelligibility index (SII) (fixed target masking sound ratio TMR), poor masking comfort and high resource consumption in the existing sound masking technology, such as Figure 3 As shown, the sound masking process may include:

[0122] (1) Collect the speech clip, and then execute step (2).

[0123] The speech segment is a speech segment with a short frame length, where the short frame length refers to a frame length less than 100 ms.

[0124] Preferably, the frame length of the voice segments collected in the present application is 40ms.

[0125] It should be noted that the frame length of the voice clips collected by the existing sound masking technology is between 100ms and 500ms. Based on the existing sound masking technology, the output of the masking signal will have a high delay and poor real-time performance, which cannot meet the low-latency sound masking requirements of the vehicle. If the frame length of the collected voice clips is less than 100ms, the existing sound masking technology will cause the output of the masking signal to have a high delay and poor real-time performance, which cannot meet the low-latency sound masking requirements of the vehicle.

[0126] By adopting the sound masking processing of the embodiment of the present application, it is possible to output a masking sound with strong masking properties in a speech segment with a short frame length while meeting the low-latency sound masking requirement of the vehicle.

[0127] (2) Perform time flipping on the speech segment, and then execute step (3).

[0128] Among them, through time reversal processing, a preliminary human voice-like masking signal can be obtained.

[0129] (3) Perform feature filtering on the speech segment after time flipping, and then execute step (4).

[0130] Among them, through feature filtering processing, the correlation between the speech segment after time reversal processing and the original speech segment is enhanced, thereby improving the masking performance.

[0131] (4) Perform spectral cropping on the speech segment after feature filtering, and then execute step (5).

[0132] Among them, through spectrum clipping processing, redundant features of the speaker can be removed, reducing the semantic comprehensibility of the speech segment itself. At the same time, through spectrum clipping processing, the part of the spectrum that exceeds the power threshold is retained and enhanced, and the part of the spectrum that is below the power threshold is suppressed, solving the problem of insufficient masking effect caused by short frame length time flip signal.

[0133] (5) Perform frequency domain noise addition processing on the speech segment after the spectrum cropping processing, and then execute step (6).

[0134] Among them, the frequency domain noise used in the embodiment of the present application is multiplicative Gaussian white noise.

[0135] By superimposing frequency domain noise in the speech segment, the semantic intelligibility of the speech segment itself is further reduced.

[0136] (6) Perform time domain noise addition processing on the speech segment after the frequency domain noise addition processing, and then execute step (7).

[0137] Among them, the time domain noise used in the embodiment of the present application is continuous steady-state noise in nature, and the weight ratio of the signal corresponding to the voice data of the voice segment in the time domain and the continuous steady-state noise in nature when superimposed is 1:1.

[0138] By superimposing temporal noise in the speech segment, the semantic intelligibility of the speech segment itself is further reduced.

[0139] (7) Perform gain control processing on the speech segment after the time domain noise addition processing, and then execute step (8).

[0140] (8) Perform windowing and smoothing processing on the speech segment after gain control processing and output the masking signal.

[0141] Among them, in this specific embodiment, if Figure 4 As shown, the specific feature filtering process in step (3) may include: the speech segment after time flipping is a time flipping signal. The time flipping signal obtained after time flipping is subjected to cepstrum analysis to extract the fundamental frequency and resonance peak characteristics of the speaker. Further, a feature filter is generated according to the fundamental frequency and resonance peak characteristics. In addition, it is also necessary to perform amplitude spectrum analysis on the speech segment after time flipping to obtain an amplitude spectrum. The speech segment after time flipping is subjected to feature filtering through the generated feature filter and amplitude spectrum, and then step (4) is executed.

[0142] Among them, in this specific embodiment, if Figure 5 As shown, the specific spectrum clipping process in step (4) may include: the speech segment after feature filtering is a feature filtering signal. The power spectrum is analyzed according to the feature filtering signal obtained after feature filtering to obtain a power spectrum. By setting a spectrum clipping function, and performing spectrum clipping on the power spectrum of the speech segment after feature filtering according to the spectrum clipping function, the part of the power spectrum exceeding the power threshold is retained and enhanced, and the part of the power spectrum below the power threshold is suppressed.

[0143] The frame length of the input voice segment of the masking signal generation algorithm module in this specific embodiment only needs 40ms, which is much shorter than the time flip frame length of 100ms~500ms in the prior art, and meets the requirements of low-delay real-time sound masking of the vehicle in the embodiment of this application. In order to further enhance the correlation between the masking sound and the target speech and improve the masking performance, the feature filtering technology of the embodiment of this application is adopted, and the speaker's fundamental frequency and formant characteristics are extracted through cepstrum analysis, and a feature filter is generated to perform feature filtering on the speech amplitude spectrum to strengthen the correlation between the masking signal and the speaker's glottal impulse and vocal tract response characteristics. In order to make up for the problem of insufficient masking effect caused by the short-frame long-time flip signal, the spectrum clipping technology of the embodiment of this application is adopted to clip the power spectrum of the masking signal frame, retain and enhance the part of the spectrum exceeding the power threshold, and suppress the part of the spectrum below the power threshold. A vehicle-mounted low-delay real-time sound masking method and device based on feature filtering and spectrum clipping proposed in this application can effectively reduce the delay of the masking system and reduce the speech clarity index SII under the condition of 40ms short-frame speech input, and does not require pre-storage of a large amount of reference speech data.

[0144] like Figure 6 As shown, Figure 6 This is a schematic diagram of the speech intelligibility index SII and delay parameter Delay of the masking signal output in different sound masking schemes under the condition of a fixed TMR of -8dB provided by the present application. Figure 6 It can be seen that under the condition of a fixed TMR of -8dB, the SII value of the masking signal output in the above preferred embodiment is lower than the white noise masking sound, so its speech clarity is lower. Correspondingly, it has a better masking effect, and because the frame length selected in the preferred embodiment is 40ms, the system delay of its masking signal generation module is 40ms, which is much lower than the 100ms~500ms delay of the long frame time flip method, and has better real-time performance.

[0145] In order to better implement the sound masking processing method provided in the embodiment of the present application, the embodiment of the present application also provides a device based on the above sound masking processing method. The meanings of the terms are the same as those in the above sound masking processing method, and the specific implementation details can refer to the description in the method embodiment.

[0146] For example, Figure 7 As shown, the sound masking processing device may include a voice collection module 201, a masking signal generation module 202 and a voice playback module 203, as follows:

[0147] The voice collection module 201 is used to obtain the target voice segment;

[0148] The masking signal generating module 202 is used to perform sound masking processing on at least one speech feature of the target speech segment to obtain a masking signal;

[0149] The voice playing module 203 is used to output a masking signal.

[0150] In the embodiment of the present application, a target voice segment is acquired through a voice acquisition module 201; a masking signal generation module 202 performs sound masking processing on at least one voice feature of the target voice segment to obtain a masking signal; and a voice playback module 203 outputs a masking signal, which can improve the user experience when performing sound masking in different vehicle scenarios.

[0151] Optionally, in an embodiment of the present application, the speech feature includes a power spectrum, and the masking signal generating module 202 includes:

[0152] A power spectrum analysis unit, used for performing power spectrum analysis on the target speech segment to obtain a power spectrum;

[0153] A spectrum clipping unit, used for performing spectrum clipping processing based on the power spectrum to obtain a clipped power spectrum;

[0154] The mask signal first determination unit is used to obtain the mask signal based on the clipped power spectrum.

[0155] Optionally, in an embodiment of the present application, the spectrum clipping unit includes:

[0156] A segment division subunit is used to segment the power spectrum according to the amplitude value in the power spectrum to obtain at least one power spectrum sub-segment, where one power spectrum sub-segment corresponds to one amplitude range;

[0157] The spectrum clipping subunit is used to clip the power spectrum sub-segments whose amplitude range is smaller than the preset amplitude value in the power spectrum to obtain the clipped power spectrum.

[0158] Optionally, in an embodiment of the present application, the speech feature includes phase information, and the masking signal generating module 202 includes:

[0159] A phase information acquisition unit, used to acquire phase information of a target speech segment;

[0160] The second masking signal determining unit is used to add phase noise to the target speech segment based on the phase information to obtain the masking signal.

[0161] Optionally, in an embodiment of the present application, before the masking signal generating module 202, it further includes:

[0162] A time information determination unit, used for performing time flip processing on the target speech segment to obtain a flipped speech segment;

[0163] The masking signal generating module 202 comprises:

[0164] The masking signal third determination unit is used to perform sound masking processing on at least one speech feature of the flipped speech segment to obtain a masking signal.

[0165] Optionally, in an embodiment of the present application, before the masking signal generating module 202, it further includes:

[0166] A feature filtering unit, used to perform feature filtering on the target speech segment to obtain a filtered speech segment;

[0167] The masking signal generating module 202 comprises:

[0168] The masking signal fourth determination unit is used to perform sound masking processing on at least one speech feature of the filtered speech segment to obtain a masking signal.

[0169] Optionally, in an embodiment of the present application, the feature filtering unit includes:

[0170] The amplitude spectrum analysis subunit is used to perform amplitude spectrum analysis on the target speech segment to obtain an amplitude spectrum;

[0171] A feature filter generation subunit is used to extract features of the target speech segment and generate a feature filter according to the extracted features;

[0172] The feature filtering subunit is used to perform feature filtering processing on the target speech segment based on the feature filter and the amplitude spectrum to obtain a filtered speech segment.

[0173] Optionally, in an embodiment of the present application, the feature filter generating subunit includes:

[0174] A cepstrum analysis subunit, used for performing cepstrum analysis on the target speech segment to obtain a fundamental frequency and at least one formant feature;

[0175] The generating subunit is used to generate a characteristic filter according to the fundamental frequency and the resonance peak characteristics.

[0176] Optionally, in an embodiment of the present application, the mask signal first determination unit includes:

[0177] A processed speech segment acquisition unit, configured to obtain a cropped processed speech segment based on the cropped power spectrum;

[0178] A noise adding unit, used for performing noise adding processing on the clipped processed speech segment to obtain a processed speech segment after noise addition;

[0179] The masking signal fifth determination unit is used to obtain the masking signal based on the processed speech segment after the noise is added.

[0180] Optionally, in an embodiment of the present application, the added noise includes at least one of frequency domain noise and time domain noise;

[0181] The noise adding unit comprises:

[0182] The first noise adding subunit is used for performing spectrum analysis on the clipped processed voice segment to obtain a spectrum when the added noise is frequency domain noise, and performing frequency domain noise adding processing on each voice data in the clipped processed voice segment based on the spectrum and the frequency domain noise to obtain a processed voice segment after noise addition;

[0183] The second noise adding subunit is used for, when the added noise is time domain noise, performing time domain noise adding processing on each voice data in the clipped processed voice segment in the time domain based on the time domain noise to obtain the processed voice segment after noise addition.

[0184] Optionally, in an embodiment of the present application, the fifth masking signal determining unit includes:

[0185] A preset target masking ratio obtaining subunit, used to obtain a preset target masking ratio;

[0186] A gain factor determination subunit, used to determine the gain factor of each speech data in the processed speech segment after adding noise based on a preset target masking ratio;

[0187] A gain control subunit, configured to perform gain control on the processed speech segment after noise addition based on the gain factor of each speech data, so as to obtain a gain speech segment;

[0188] The masking signal fifth determination subunit is used to obtain the masking signal based on the gain speech segment.

[0189] Optionally, in an embodiment of the present application, the voice collection module 201 includes:

[0190] A voice collection unit, used to obtain the voice uttered by the user in the first space through a voice collection module set in the first space of the vehicle, and obtain a target voice segment;

[0191] The above-mentioned voice playing module 203 includes:

[0192] The voice playing unit is used to output a masking signal to the second space through a voice playing module arranged in the second space of the vehicle.

[0193] In the sound masking processing device of the embodiment of the present application, the target speech segment is acquired through the speech acquisition module 201; the masking signal generation module 202 performs sound masking processing on at least one speech feature of the target speech segment to obtain a masking signal; and the speech playback module 203 outputs the masking signal. By performing sound masking processing on at least one speech feature of the target speech segment, the obtained masking signal meets the low delay requirement and improves the user experience.

[0194] In specific implementation, the above modules can be implemented as independent entities, or can be arbitrarily combined and implemented as the same or several entities. The specific implementation methods and corresponding beneficial effects of the above modules can be found in the previous method embodiments, which will not be repeated here.

[0195] An embodiment of the present application further provides a vehicle, which includes a first space and a second space. A voice collection module is provided in the first space, a voice playback module is provided in the second space, and a masking signal generation module is also provided on the vehicle.

[0196] In the vehicle, various functions are implemented, such as:

[0197] A voice collection module is provided in the first space of the vehicle to obtain a target voice segment;

[0198] The masking signal generating module provided by the vehicle is used to perform sound masking processing on at least one speech feature of the target speech segment to obtain a masking signal;

[0199] The voice playing module arranged in the second space of the vehicle is used to output the masking signal.

[0200] The vehicle provided in the embodiment of the present application is used to obtain a target voice segment through a voice acquisition module set in the first space of the vehicle; a masking signal generation module set in the vehicle is used to perform sound masking processing on at least one voice feature of the target voice segment to obtain a masking signal; and a voice playback module set in the second space of the vehicle is used to output the masking signal. Based on this, by performing sound masking processing on at least one voice feature of the target voice segment, the obtained masking signal meets the low latency requirement and improves the user experience.

[0201] like Figure 2 As shown, Figure 2 It is a structural schematic diagram of a vehicle carrier provided in an embodiment of the present application. Figure 2 The seating area in the rear cabin is the privacy protection area, which specifically refers to the seating area in the car that is protected by voice privacy, that is, the area where private voice is generated, that is, the first space. Figure 2The front cockpit seat area is the voice isolation area, which specifically refers to the seat area in the car where private voices are not expected to be leaked, that is, the area where interference signals are masked, that is, the second space. Figure 2 The first space shown is equipped with a microphone, namely, a voice collection module. Figure 2 The second space shown is equipped with a headrest speaker, i.e., a voice playback module. Figure 2 The vehicle-mounted computing platform shown is configured with a masking signal generating module.

[0202] The voice to be masked in the privacy protection zone is collected by the microphone inside the car and transmitted in frames to the masking signal generation module in the form of digital signals. The masking signal is calculated and output by the on-board computing platform, and finally played by the headrest speakers in the voice isolation zone, thereby interfering with the voice reception of passengers in the voice isolation zone.

[0203] like Figure 8 As shown, it shows a schematic diagram of the structure of the masking signal generating module involved in the embodiment of the present application, specifically:

[0204] The masking signal generating module may include one or more processors 301 of processing cores, one or more computer-readable storage media memories 302, a power supply 303, an input unit 304 and other components. Those skilled in the art will appreciate that Figure 8 The structure of the masking signal generating module shown in the figure does not constitute a limitation on the masking signal generating module, and may include more or less components than shown in the figure, or combine certain components, or arrange the components differently.

[0205] The processor 301 is the control center of the masking signal generation module. It uses various interfaces and lines to connect the various parts of the entire masking signal generation module. It executes various functions and processes data of the masking signal generation module by running or executing computer programs and / or modules stored in the memory 302, and calling data stored in the memory 302. Optionally, the processor 301 may include one or more processing cores; preferably, the processor 301 may integrate an application processor and a modem processor, wherein the application processor mainly processes the operating system, user interface, and application programs, and the modem processor mainly processes wireless communications. It is understandable that the above-mentioned modem processor may not be integrated into the processor 301.

[0206] The memory 302 can be used to store computer programs and modules. The processor 301 executes various functional applications and data processing by running the computer programs and modules stored in the memory 302. The memory 302 can mainly include a program storage area and a data storage area, wherein the program storage area can store an operating system, a computer program required for at least one function (such as a sound playback function, an image playback function, etc.), etc.; the data storage area can store data created according to the use of the masking signal generation module, etc. In addition, the memory 302 can include a high-speed random access memory, and can also include a non-volatile memory, such as at least one disk storage device, a flash memory device, or other volatile solid-state storage devices. Accordingly, the memory 302 can also include a memory controller to provide the processor 301 with access to the memory 302.

[0207] The masking signal generating module also includes a power supply 303 for supplying power to each component. Preferably, the power supply 303 can be logically connected to the processor 301 through a power management system, so as to realize functions such as management of charging, discharging, and power consumption management through the power management system. The power supply 303 can also include any components such as one or more DC or AC power supplies, recharging systems, power failure detection circuits, power converters or inverters, and power status indicators.

[0208] The masking signal generating module may further include an input unit 304, which may be used to receive input digital or character information and generate keyboard, mouse, joystick, optical or trackball signal inputs related to user settings and function controls.

[0209] Although not shown, the masking signal generating module may further include a display unit, etc., which will not be described in detail herein. Specifically in this embodiment, the processor 301 in the masking signal generating module will load the executable files corresponding to the processes of one or more computer programs into the memory 302 according to the following instructions, and the processor 301 will run the computer programs stored in the memory 302, thereby realizing various functions, such as:

[0210] Get the target speech segment;

[0211] Performing sound masking processing on at least one speech feature of the target speech segment to obtain a masking signal;

[0212] The mask signal is output.

[0213] It can be seen that the masking signal generation module provided in the embodiment of the present application obtains a target voice segment; performs sound masking processing on at least one voice feature of the target voice segment to obtain a masking signal; and outputs the masking signal. The sound masking processing method provided in the present application. By performing sound masking processing on at least one voice feature of the target voice segment, the user experience when performing sound masking in different vehicle scenarios is improved.

[0214] The specific implementation methods and corresponding beneficial effects of the above operations can be found in the detailed description of the sound masking processing method above, which will not be elaborated here.

[0215] To this end, an embodiment of the present application provides a computer-readable storage medium, in which a computer program is stored, and the computer program can be loaded by a processor to execute the steps in any of the sound masking processing methods provided in the embodiments of the present application. For example, the computer program can execute the following steps:

[0216] Get the target speech segment;

[0217] Performing sound masking processing on at least one speech feature of the target speech segment to obtain a masking signal;

[0218] The mask signal is output.

[0219] It can be seen that the computer-readable storage medium provided in the embodiment of the present application obtains a target speech segment; performs sound masking processing on at least one speech feature of the target speech segment to obtain a masking signal; and outputs the masking signal. Based on this, by performing sound masking processing on at least one speech feature of the target speech segment, the obtained masking signal meets the low latency requirement and improves the user experience.

[0220] The specific implementation methods and corresponding beneficial effects of the above operations can be found in the previous embodiments, which will not be described in detail here.

[0221] The computer-readable storage medium may include: a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk, etc.

[0222] Since the computer program stored in the computer-readable storage medium can execute the steps in any one of the sound masking processing methods provided in the embodiments of the present application, the beneficial effects that can be achieved by any one of the sound masking processing methods provided in the embodiments of the present application can be achieved. For details, please refer to the previous embodiments and will not be repeated here.

[0223] According to one aspect of the present application, a computer program product or a computer program is provided, the computer program product or the computer program includes computer instructions, the computer instructions are stored in a computer-readable storage medium. A processor of a computer device reads the computer instructions from the computer-readable storage medium, and the processor executes the computer instructions, so that the computer device performs the above-mentioned sound masking processing method.

[0224] The above is a detailed introduction to a sound masking processing method, device, carrier and computer-readable storage medium provided in the embodiments of the present application. Specific examples are used herein to illustrate the principles and implementation methods of the present application. The description of the above embodiments is only used to help understand the method and core idea of ​​the present application. At the same time, for those skilled in the art, according to the idea of ​​the present application, there will be changes in the specific implementation method and application scope. In summary, the content of this specification should not be understood as limiting the present application.

Claims

1. A sound masking processing method, characterized in that: include: Get the target speech segment; Performing sound masking processing on at least one speech feature of the target speech segment to obtain a masking signal; The sound masking process includes: performing power spectrum analysis on the target speech segment to obtain a power spectrum; segmenting the power spectrum according to the amplitude value in the power spectrum to obtain at least one power spectrum sub-segment, wherein one power spectrum sub-segment corresponds to an amplitude range; in the power spectrum, cutting the power spectrum sub-segment whose amplitude range is less than a preset amplitude value to obtain a cut power spectrum; and obtaining a masking signal based on the cut power spectrum; The mask signal is output.

2. The sound masking processing method according to claim 1, characterized in that: The speech feature includes phase information, and performing sound masking processing on at least one speech feature of the target speech segment to obtain a masking signal further includes: Acquiring phase information of the target speech segment; Phase noise is added to the target speech segment based on the phase information to obtain a masking signal.

3. The sound masking processing method according to claim 1, characterized in that: Before performing sound masking processing on at least one speech feature of the target speech segment to obtain a masking signal, the method further includes: Performing time flipping processing on the target speech segment to obtain a flipped speech segment; The step of performing sound masking processing on at least one speech feature of the target speech segment to obtain a masking signal includes: Perform sound masking processing on at least one speech feature of the flipped speech segment to obtain a masking signal.

4. The sound masking processing method according to claim 1, characterized in that: Before performing sound masking processing on at least one speech feature of the target speech segment to obtain a masking signal, the method further includes: Performing feature filtering on the target speech segment to obtain a filtered speech segment; The step of performing sound masking processing on at least one speech feature of the target speech segment to obtain a masking signal includes: Performing sound masking processing on at least one speech feature of the filtered speech segment to obtain a masking signal.

5. The sound masking processing method according to claim 4, characterized in that: The step of performing feature filtering on the target speech segment to obtain a filtered speech segment includes: Performing amplitude spectrum analysis on the target speech segment to obtain an amplitude spectrum; Extracting features from the target speech segment and generating a feature filter based on the extracted features; Based on the characteristic filter and the amplitude spectrum, the target speech segment is subjected to characteristic filtering processing to obtain a filtered speech segment.

6. The sound masking processing method according to claim 5, characterized in that: The step of extracting features from the target speech segment and generating a feature filter according to the extracted features includes: Performing cepstrum analysis on the target speech segment to obtain a fundamental frequency and at least one formant feature; A characteristic filter is generated according to the fundamental frequency and the resonance peak characteristics.

7. The sound masking processing method according to claim 1, characterized in that: The step of obtaining a masking signal based on the clipped power spectrum comprises: Based on the clipped power spectrum, obtaining a clipped processed speech segment; Performing noise adding processing on the clipped processed speech segment to obtain a processed speech segment with noise added; A masking signal is obtained based on the processed speech segment after the noise is added.

8. The sound masking processing method according to claim 7, characterized in that: The added noise includes at least one of frequency domain noise and time domain noise; The step of performing noise addition processing on the clipped processed speech segment to obtain the processed speech segment after noise addition comprises: When the added noise is frequency domain noise, performing spectrum analysis on the clipped processed speech segment to obtain a spectrum, and based on the spectrum and the frequency domain noise, performing frequency domain noise addition processing on each speech data in the clipped processed speech segment to obtain a processed speech segment to which noise is added; When the added noise is time domain noise, based on the time domain noise, each voice data in the clipped processed voice segment in the time domain is subjected to time domain noise addition processing to obtain a processed voice segment with added noise.

9. The sound masking processing method according to claim 7, characterized in that: The step of obtaining a masking signal based on the processed speech segment after the noise is added comprises: Get the preset target masking ratio; Based on the preset target masking ratio, determining a gain factor of each speech data in the processed speech segment after adding noise; Based on the gain factor of each of the voice data, gain control is performed on the processed voice segment after the noise is added to obtain a gain voice segment; Based on the gain speech segment, a masking signal is obtained.

10. The sound masking processing method according to any one of claims 1 to 9, characterized in that: The step of obtaining the target voice segment comprises: The voice acquisition module provided in the first space of the vehicle is used to acquire the voice uttered by the user in the first space to obtain the target voice segment; The outputting the masking signal comprises: The masking signal is output to the second space through a voice playing module arranged in the second space of the vehicle.

11. A sound masking processing device, characterized in that: include: A voice collection module is used to obtain target voice segments; The masking signal generating module is used to perform sound masking processing on at least one speech feature of the target speech segment to obtain a masking signal; the sound masking processing includes: performing power spectrum analysis on the target speech segment to obtain a power spectrum; segmenting the power spectrum according to the amplitude value in the power spectrum to obtain at least one power spectrum sub-segment, wherein one power spectrum sub-segment corresponds to an amplitude range; in the power spectrum, cutting the power spectrum sub-segment whose amplitude range is less than a preset amplitude value to obtain a cut power spectrum; and obtaining the masking signal based on the cut power spectrum; The voice playing module is used to output the masking signal.

12. A carrier, characterized in that: The vehicle comprises a first space and a second space, the first space is provided with a voice collection module as claimed in claim 11, the second space is provided with a voice playback module as claimed in claim 11, and the vehicle is also provided with a masking signal generation module as claimed in claim 11.

13. The carrier according to claim 12, characterized in that: The vehicle includes at least one of a vehicle vehicle, a flying vehicle, a motor vehicle vehicle and a ship vehicle.

14. A computer-readable storage medium, characterized in that: The device comprises a computer program. When the computer program is run on a vehicle, the computer program is used to enable the vehicle to execute the steps of the sound masking processing method according to any one of claims 1 to 10.

Images