Electronic equipment voice signal processing method and system

Abstract

The invention relates to the technical field of voice data processing, in particular to a voice signal processing method and system for electronic equipment. The method comprises the following steps: converting a collected voice signal into a frequency domain signal; dividing a silent segment and a voice segment in the voice signal into a plurality of overlapped short-time frames, and calculating voice signal characteristic values of the short-time frames; calculating the affected amplitude of the target short-time frame; calculating a noise adjustment parameter of the target short-time frame; determining a frequency spectrum after noise reduction in the frequency domain signal of the target short time frame; and according to the frequency spectrum after noise reduction, realizing voice signal processing of the electronic equipment. According to the method, the noise is flexibly adjusted by calculating the affected amplitude and the noise adjustment parameter of the target short-time frame, the voice signal distortion is avoided, and the noise reduction effect is improved; noise estimation is dynamically optimized in combination with voice signal feature values of neighborhood segments, and the noise reduction accuracy is improved; by adjusting the frequency spectrum after noise reduction, the real tone quality of the voice signal is reserved, and the voice definition and intelligibility are enhanced.

Description

Technical Field

[0001] The present invention relates to the technical field of voice data processing, and in particular to a method and system for processing voice signals of electronic equipment. Background Art

[0002] In a noisy environment, voice signal processing technology can be used to distinguish voice signals from background noise, improve the signal-to-noise ratio of voice signals, and thus improve call quality. In mixed voice scenarios such as multi-person meetings or telephone conferences, voice separation technology can separate the voice signals of different speakers from mixed voice signals, which is essential for applications such as voice recognition, automatic transcription, and conference recording. At the same time, voice enhancement is performed, and spectral subtraction is used to remove noise and improve the clarity of voice signals, which is very important for improving the efficiency and quality of voice communication. It is also convenient to post-process voice signals, such as voice coding and compression, to facilitate storage and transmission. In addition, through pre-processing technologies such as pre-emphasis, framing and windowing, and feature parameter extraction, voice endpoint detection is achieved, which is essential for the effective processing and analysis of voice signals. Voice noise reduction can also be performed, and spectral subtraction and other technologies are used to reduce noise in voice signals and improve voice clarity and intelligibility.

[0003] The patent application document with publication number CN116343815A discloses a speech enhancement method based on Wiener filtering with a priori signal-to-noise ratio and improved spectral subtraction. The method first requires an external input noisy speech signal, then the noisy speech signal is windowed into frames and Fourier transformed to obtain a noisy speech signal in the frequency domain, then the speech signal is preliminarily processed by Wiener filtering based on a priori signal-to-noise ratio, and finally the processed speech signal is further processed by improved spectral subtraction and inverse Fourier transformed to obtain a relatively pure speech signal.

[0004] However, when using spectral subtraction for speech denoising, since spectral subtraction usually uses a noise signal in a silent period of time to denoise the speech, but the influence of noise is not always constant, the traditional spectral subtraction will cause errors in the denoised speech signal, resulting in poor noise reduction effect and unclear speech signal after noise reduction, thereby reducing the accuracy of the speech signal. Summary of the invention

[0005] In order to solve the problem that the influence of noise is not always constant, the traditional spectral subtraction method will cause errors in the noise-reduced speech signal, resulting in poor noise reduction effect and unclear speech signal after noise reduction, thereby reducing the accuracy of the speech signal, the present invention provides a method and system for processing speech signals of an electronic device.

[0006] In a first aspect, the present invention provides a method for processing a speech signal of an electronic device, which adopts the following technical solution: A method for processing speech signals of electronic equipment comprises: converting a collected speech signal into a frequency domain signal; dividing the silence segment and speech segment in the speech signal into a plurality of overlapping short time frames respectively; for each short time frame, calculating the speech signal characteristic value of the short time frame based on the average amplitude in the speech signal of the short time frame, the average frequency and the average power in the frequency domain signal of the short time frame; recording any speech segment as a target segment, recording any short time frame in the target segment as a target short time frame, and recording all adjacent silence segments of the target segment as neighborhood segments; calculating the affected amplitude of the target short time frame based on the speech signal characteristic values ​​of all short time frames in the target segment and the average of the speech signal characteristic values ​​of all short time frames in the neighborhood segments; calculating the noise adjustment parameter of the target short time frame based on the affected amplitudes of the target short time frame and all adjacent short time frames of the target short time frame in the target segment and the average of the speech signal characteristic values ​​of all short time frames in the neighborhood segments; determining the noise-reduced spectrum in the frequency domain signal of the target short time frame, , For short time frame The denoised spectrum in the frequency domain signal is and Short time frame The noise adjustment parameter and the spectrum of the frequency domain signal, For short time frame The average spectrum of the neighboring segments of the speech segment, is a maximum function; and based on the noise-reduced spectrum, speech signal processing of the electronic device is realized.

[0007] By calculating the affected amplitude and noise adjustment parameters of the target short-time frame, the noise can be flexibly adjusted to avoid speech signal distortion and improve the noise reduction effect. By combining the speech signal feature values ​​of the neighborhood segment, the noise estimation is dynamically optimized to improve the accuracy of noise reduction and avoid excessive noise reduction or speech distortion. By adjusting the spectrum after noise reduction, the true sound quality of the speech signal is retained and the clarity and intelligibility of the speech are enhanced.

[0008] Furthermore, the essence of the speech signal is a time domain signal, and the frequency domain signal is obtained by converting the time domain signal into a frequency domain signal by using Fourier transform.

[0009] Furthermore, the speech signal characteristic value satisfies the following relationship: ; In the formula, For short time frame The speech signal feature value, For short time frame The average amplitude in the speech signal, and Short time frame The average frequency and average power in the frequency domain signal.

[0010] Combining the average amplitude in the time domain, the average frequency in the frequency domain, and the average power into a characteristic value provides a more comprehensive and accurate description of the speech signal, which helps to improve the subsequent algorithm's ability to recognize and process speech signals; by comprehensively considering the time domain and frequency domain signals, the speech signal characteristic value can more effectively distinguish speech from noise, thereby optimizing the noise reduction effect and reducing noise mis-suppression and speech distortion.

[0011] Furthermore, the affected amplitude satisfies the following relationship: ; In the formula, For short time frame The extent of the impact, For short time frame The speech signal feature value, For short time frame The mean of the speech signal feature values ​​of all short-time frames in the neighborhood segment of the speech segment to which it belongs, For short time frame The first The speech signal feature value of a short time frame, For short time frame The number of short-time frames in the speech segment. is the absolute value symbol.

[0012] By comparing the feature means of the short-time frame and its neighboring segments, the part of the signal that is most affected by noise can be identified, so that stronger noise reduction processing can be carried out in a targeted manner; according to the affected amplitude of each frame, the noise reduction intensity can be adaptively adjusted to optimize the noise reduction strategy; by calculating the affected amplitude, the noise reduction algorithm can dynamically adjust the processing strategy to enhance its adaptability and stability in different noise environments, thereby improving the overall noise reduction effect.

[0013] Furthermore, the noise adjustment parameter satisfies the following relationship: ; In the formula, For short time frame and short time frames within the speech segment Among all adjacent short time frames The magnitude of the impact of a short time frame, For short time frame The mean of the speech signal feature values ​​of all short-time frames in the neighborhood segment of the speech segment to which it belongs, For short time frame and short time frames within the speech segment The number of all adjacent short-time frames, is the normalization function.

[0014] By comprehensively considering the neighborhood information of short-time frames, the noise suppression strength is dynamically adjusted to effectively optimize the processing for different noise environments; adaptive adjustments are made according to the degree to which each frame is affected by noise to avoid excessive noise reduction and maintain the naturalness of the speech signal; normalization and local information weighting enable the noise reduction algorithm to adapt to different noise environments and improve overall robustness and performance.

[0015] Furthermore, the average spectrum of the neighborhood segment is the average of the spectrums of all short-time frames in the neighborhood segment.

[0016] Furthermore, the voice signal processing of the electronic device includes: converting the noise-reduced spectrum back into a voice signal using an inverse Fourier transform to obtain a noise-reduced voice signal, overlapping and adding all noise-reduced voice signals to obtain a complete noise-reduced voice signal, and completing the voice signal processing of the electronic device.

[0017] The denoised spectrum is converted into a speech signal through inverse Fourier transform, effectively restoring the clarity and naturalness of the speech signal; after denoising in the frequency domain, overlap-addition processing is performed to reduce residual noise and improve the intelligibility of the speech signal; it can process speech signals in different noise environments and enhance the robustness of the system.

[0018] In a second aspect, the present invention provides a voice signal processing system for an electronic device, which adopts the following technical solution: A speech signal processing system for electronic equipment comprises: a processor and a memory, wherein the memory stores computer program instructions, and when the computer program instructions are executed by the processor, the above-mentioned speech signal processing method for electronic equipment is implemented.

[0019] By adopting the above technical solution, the above-mentioned electronic device voice signal processing method is generated into a computer program and stored in a memory so as to be loaded and executed by a processor, thereby making a terminal device based on the memory and the processor for easy use.

[0020] The present invention has the following technical effects: The adjustment parameters of each short-time frame are obtained through the speech signal characteristic values ​​of the silent segment and the speech segment, and each short-time frame is adaptively adjusted according to the adjustment parameters. Finally, the speech signal processing of the electronic device is realized by using the denoised spectrum, which ensures the effectiveness of the noise reduction effect, reduces the noise in the speech signal, and improves the clarity and intelligibility of the speech; at the same time, it can adapt to different speech and noise environments to ensure that a good noise reduction effect can be maintained in various scenarios. By comprehensively analyzing the time domain and frequency domain characteristics of the short-time frame, the error caused by noise is reduced and the noise reduction quality is improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] By reading the detailed description below with reference to the accompanying drawings, the above and other purposes, features and advantages of the exemplary embodiments of the present invention will become readily understood. In the accompanying drawings, several embodiments of the present invention are shown in an exemplary and non-limiting manner, and the same or the same reference numerals are the same or the same parts.

[0022] Figure 1 The present invention is a flowchart of a method for processing a speech signal of an electronic device according to an embodiment of the present invention. DETAILED DESCRIPTION

[0023] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative work are within the scope of protection of the present invention.

[0024] It should be understood that when the terms "first", "second", etc. are used in the claims, descriptions, and drawings of the present invention, they are only used to distinguish different objects, rather than to describe a specific order. The terms "include" and "comprise" used in the description and claims of the present invention indicate the presence of the described features, wholes, steps, operations, elements, and / or components, but do not exclude the presence or addition of one or more other features, wholes, steps, operations, elements, components, and / or their collections.

[0025] The embodiment of the present invention discloses a method for processing a speech signal of an electronic device, referring to Figure 1 , comprising steps S1 to S6: S1: Convert the collected speech signal into a frequency domain signal.

[0026] Specifically, the essence of the speech signal is a time domain signal, and the frequency domain signal is obtained by converting the time domain signal into a frequency domain signal by using Fourier transform.

[0027] S2: Divide the silence segments and speech segments in the speech signal into several overlapping short-time frames, and calculate the speech signal feature values ​​of the short-time frames.

[0028] The silence segments and speech segments in the speech signal are divided into several overlapping short-time frames.

[0029] Implementers can set the frame length and frame shift of the short-time frame according to specific implementation conditions. For example, the frame length ranges from 20 to 30 milliseconds, and the frame shift is half of the frame length.

[0030] It should be noted that after obtaining each short-time frame, the characteristics of the speech signal data are comprehensively analyzed using the information of the speech signal and the frequency domain signal corresponding to each short-time frame, which facilitates subsequent analysis and reduces the amount of calculation.

[0031] For each short time frame, a speech signal feature value of the short time frame is calculated based on the average amplitude in the speech signal of the short time frame, the average frequency and the average power in the frequency domain signal of the short time frame.

[0032] Specifically, the speech signal characteristic value satisfies the following relationship: ; In the formula, For short time frame The speech signal feature value, For short time frame The average amplitude in the speech signal, and Short time frame The average frequency and average power in the frequency domain signal.

[0033] S3: Calculate the affected amplitude of the target short-time frame.

[0034] It should be noted that in electronic speech signal processing, the above steps are used to divide the signal into multiple short-time frames for analysis. Ideally, the noise impact on each short-time frame should be approximately the same. However, due to the uncertainty and randomness of noise, the impact of noise will be different even in consecutive short-time frames, which will cause the speech signal characteristics that should have been stable to become unstable and incoherent. Therefore, it is necessary to calculate the affected amplitude of each short-time frame.

[0035] Any speech segment is recorded as a target segment, any short-time frame in the target segment is recorded as a target short-time frame, and all adjacent silent segments of the target segment are recorded as neighborhood segments; based on the speech signal feature values ​​of all short-time frames in the target segment and the average of the speech signal feature values ​​of all short-time frames in the neighborhood segments, the affected amplitude of the target short-time frame is calculated; all adjacent silent segments are silent segments adjacent to the target segment. If there are adjacent silent segments on both sides of the target segment, there are two adjacent silent segments. If there is an adjacent silent segment on only one side, there is one adjacent silent segment.

[0036] Specifically, the affected amplitude satisfies the following relationship: ; In the formula, For short time frame The extent of the impact, For short time frame The speech signal feature value, For short time frame The mean of the speech signal feature values ​​of all short-time frames in the neighborhood segment of the speech segment to which it belongs, For short time frame The first The speech signal feature value of a short time frame, For short time frame The number of short-time frames in the speech segment. is the absolute value symbol.

[0037] in, Represents a short time frame The larger the value, the more the short-time frame is affected by noise. Represents a short time frame The mean of the speech signal feature values ​​of all short-time frames in the neighborhood segment of the speech segment is used to reflect the characteristics of the noise signal; Reflecting short time frame In eliminating the normal features of the predictions affected by noise, Reflecting short time frame The first The normal characteristics of the prediction of a short time frame in eliminating the noise influence are obtained by Reflects the short time frame after eliminating the noise effect and short time frame The similarity between .

[0038] S4: Calculate the noise adjustment parameters of the target short-time frame.

[0039] It should be noted that in the noise reduction process, it is necessary not only to consider the noise impact amplitude of each short-time frame, but also to adjust the strength of the noise reduction algorithm in combination with the characteristics of adjacent short-time frames and silent segments; by finely adjusting the affected amplitude of each short-time frame and obtaining the noise adjustment parameters of each short-time frame, it is possible to more accurately suppress the noise components while maintaining the effective components of the speech signal.

[0040] The noise adjustment parameter of the target short-time frame is calculated based on the affected amplitudes of the target short-time frame and all adjacent short-time frames of the target short-time frame in the target segment, and the mean of the speech signal characteristic values ​​of all short-time frames in the neighborhood segment; all adjacent short-time frames of the target short-time frame in the target segment are short-time frames adjacent to the target short-time frame in the same speech segment. If there are adjacent short-time frames on both sides of the target, there are two adjacent short-time frames. If there are adjacent short-time frames on only one side, there is one adjacent short-time frame.

[0041] Specifically, the noise adjustment parameter satisfies the following relationship: ; In the formula, For short time frame The noise adjustment parameter, For short time frame and short time frames within the speech segment Among all adjacent short time frames The magnitude of the impact of a short time frame, For short time frame The mean of the speech signal feature values ​​of all short-time frames in the neighborhood segment of the speech segment to which it belongs, For short time frame and short time frames within the speech segment The number of all adjacent short-time frames, is the normalization function.

[0042] in, It directly reflects the degree to which the target short-time frame and its adjacent short-time frames are affected by noise. A larger value indicates that the short-time frame is seriously affected by noise. When increases, resulting in an increase in the noise suppression strength of the short-time frame and its adjacent frames during the noise reduction process. For example, if the short-time frame The signal of its adjacent frames fluctuates greatly in a noisy environment. The overall size is large, then the value calculated according to the formula will become larger accordingly, and these frames will be processed more strongly during noise reduction to reduce the impact of noise; It reflects the overall characteristics of the speech signal in the neighborhood segment of the speech segment to which the target short-time frame belongs. If The larger the value, the stronger the overall speech signal in the neighborhood segment. When Increase, which means that during noise reduction, the noise reduction strength of the target short-time frame and its adjacent frames will be adjusted according to the strength of the speech signal in the neighborhood segment. For example, when the strength of the speech signal in the neighborhood segment is large, in order to maintain the continuity and integrity of the speech signal, the noise reduction strength of the target short-time frame will be appropriately increased to better adapt to the overall speech environment.

[0043] S5: Determine a noise-reduced spectrum in the frequency domain signal of the target short-time frame.

[0044] The denoised spectrum in the frequency domain signal of the target short-time frame satisfies the following relationship: ; In the formula, For short time frame The denoised spectrum in the frequency domain signal is and Short time frame The noise adjustment parameter and the spectrum of the frequency domain signal, For short time frame The average spectrum of the neighboring segments of the speech segment, is the maximum value function.

[0045] in, It reflects the spectral characteristics around the speech segment to which the target short-time frame belongs. This value serves as a reference benchmark when calculating the denoised spectrum. The larger the calculation When the original spectrum The more parts are subtracted from the image, the greater the noise reduction is for the short time frame. When increasing, The value of will also increase, thereby more significantly weakening the noise component in the original spectrum. For example, in a noisy environment, the calculated may be large, in which case the influence of noise on the spectrum can be effectively reduced, making the spectrum after noise reduction closer to the spectrum of pure speech; however, if and If it is too large, it will lead to excessive noise reduction, so that some effective components of the speech signal are also mistakenly removed as noise, thus affecting the speech quality. Therefore, at this time, 0 is selected as the noise-reduced spectrum in the frequency domain signal of the short-time frame.

[0046] Specifically, the average spectrum of the neighborhood segment is the average of the spectrums of all short-time frames in the neighborhood segment.

[0047] S6: Implementing voice signal processing of the electronic device according to the noise-reduced spectrum.

[0048] Specifically, the method for implementing voice signal processing of an electronic device includes: The noise-reduced spectrum is converted back into a speech signal using inverse Fourier transform to obtain a noise-reduced speech signal, and all noise-reduced speech signals are overlapped and added to obtain a complete noise-reduced speech signal, thereby ensuring the effectiveness of the noise reduction effect, reducing the noise in the speech signal, improving the clarity and intelligibility of the speech, and completing the speech signal processing of the electronic device.

[0049] Implementers can select overlapping methods according to specific implementation methods, such as overlapping windowing method.

[0050] The embodiment of the present invention further discloses an electronic device voice signal processing system, including a processor and a memory, wherein the memory stores computer program instructions, and when the computer program instructions are executed by the processor, an electronic device voice signal processing method according to the present invention is implemented.

[0051] The above system also includes other components well known to those skilled in the art, such as a communication bus and a communication interface, and their configuration and functions are known in the art, so they will not be described in detail here.

[0052] In the present invention, the aforementioned memory may be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, apparatus, or device. For example, a computer-readable storage medium may be any suitable magnetic storage medium or magneto-optical storage medium, or any other medium that can be used to store the required information and can be accessed by an application, module, or both. Any such computer storage medium may be part of the device or accessible or connectable to the device.

[0053] Although this specification has shown and described a number of embodiments of the present invention, it will be apparent to those skilled in the art that such embodiments are provided by way of example only. Those skilled in the art will conceive of many modifications, changes and alternatives without departing from the ideas and spirit of the present invention. It should be understood that in the practice of the present invention, various alternatives to the embodiments of the present invention described herein may be employed.

[0054] The above are all preferred embodiments of the present invention, and are not intended to limit the protection scope of the present invention. Therefore, any equivalent changes made based on the structure, shape, and principle of the present invention should be included in the protection scope of the present invention.

Claims

1. A method for processing speech signals of an electronic device, characterized in that: include: Convert the collected speech signal into a frequency domain signal; The silence segment and speech segment in the speech signal are divided into a number of overlapping short time frames respectively; For each short time frame, based on the average amplitude in the speech signal of the short time frame, the average frequency and the average power in the frequency domain signal of the short time frame, a speech signal characteristic value of the short time frame is calculated; Any speech segment is recorded as a target segment, any short-time frame in the target segment is recorded as a target short-time frame, and all adjacent silent segments of the target segment are recorded as neighborhood segments; Calculate the affected amplitude of the target short-time frame based on the speech signal characteristic values ​​of all short-time frames in the target segment and the average of the speech signal characteristic values ​​of all short-time frames in the neighboring segment; Calculate the noise adjustment parameter of the target short-time frame based on the affected amplitudes of the target short-time frame and all adjacent short-time frames of the target short-time frame in the target segment, and the mean of the speech signal characteristic values ​​of all short-time frames in the neighborhood segment; Determine the noise-reduced spectrum in the frequency domain signal of the target short-time frame, , For short time frame The denoised spectrum in the frequency domain signal of and Short time frame The noise adjustment parameter and the spectrum of the frequency domain signal, For short time frame The average spectrum of the neighboring segments of the speech segment, is the maximum value function; The speech signal processing of the electronic device is realized according to the noise-reduced spectrum.

2. The method for processing speech signals of an electronic device according to claim 1, characterized in that: The essence of the speech signal is a time domain signal, and the frequency domain signal is obtained by converting the time domain signal into a frequency domain signal by using Fourier transform.

3. The method for processing speech signals of an electronic device according to claim 1, characterized in that: The speech signal characteristic value satisfies the following relationship: ； In the formula, For short time frame The speech signal feature value, For short time frame The average amplitude in the speech signal, and Short time frame The average frequency and average power in the frequency domain signal.

4. The method for processing speech signals of an electronic device according to claim 1, characterized in that: The affected amplitude satisfies the following relationship: ； In the formula, For short time frame The extent of the impact, For short time frame The speech signal feature value, For short time frame The mean of the speech signal feature values ​​of all short-time frames in the neighborhood segment of the speech segment to which it belongs, For short time frame The first The speech signal feature value of a short time frame, For short time frame The number of short-time frames in the speech segment. is the absolute value symbol.

5. The method for processing speech signals of an electronic device according to claim 1, characterized in that: The noise adjustment parameter satisfies the following relationship: ； In the formula, For short time frame and short time frames within the speech segment Among all adjacent short time frames The magnitude of the impact of a short time frame, For short time frame The mean of the speech signal feature values ​​of all short-time frames in the neighborhood segment of the speech segment to which it belongs, For short time frame and short time frames within the speech segment The number of all adjacent short-time frames, is the normalization function.

6. The method for processing speech signals of an electronic device according to claim 1, characterized in that: The average spectrum of the neighborhood segment is the average of the spectrums of all short-time frames in the neighborhood segment.

7. The method for processing speech signals of an electronic device according to claim 1, characterized in that: The method for realizing voice signal processing of electronic equipment comprises: The noise-reduced spectrum is converted back into a speech signal by using inverse Fourier transform to obtain a noise-reduced speech signal, and all noise-reduced speech signals are overlapped and added to obtain a complete noise-reduced speech signal, thereby completing the speech signal processing of the electronic device.

8. A method and system for processing speech signals of electronic equipment, characterized in that: include: A processor and a memory, wherein the memory stores computer program instructions, and when the computer program instructions are executed by the processor, an electronic device voice signal processing method according to any one of claims 1 to 7 is implemented.

Images