Noise reduction and speech enhancement method with double microphone cores
Through dual microphone and adaptive filtering technology, combined with Wiener filtering and harmonic regeneration, the problem of audio signal noise interference in complex environments is solved, and effective denoising and voice enhancement in low signal-to-noise ratio environments is achieved. In a variety of high-quality audio application scenarios.
Patent Information
- Application Number
- CN202411205038.2
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-08-30
- Publication Date
- 2024-12-03
AI Technical Summary
The audio signals collected by existing single microphones in complex environments are susceptible to environmental noise interference. The existing denoising method is not effective in environments with relatively low signal-to-noise, and an effective one is urgently needed in complex environments. Sound denoising and enhancing methods.
The dual-microphone is used to collect dual-channel sound signals, noise estimation and adaptive filtering are performed through the minimum mean square error algorithm, the Wiener filter gain is calculated, and harmonic regeneration technology is combined to remove noise and enhance voice signals.
Effectively suppress noise and enhance voice signals in complex environments, improve audio quality, and is suitable for scenes such as hearing aids, smart homes, voice recognition systems and remote meeting systems that require high-quality audio input.
Smart Images

Figure CN119068895A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of audio processing and speech signal processing, and in particular to a method for noise reduction and speech enhancement with dual microphone cores. Background Art
[0002] The audio signals collected by the existing single-microphone in complex environments are easily interfered by environmental noise, which affects the audio quality. The dual-microphone can improve the accuracy of noise estimation by collecting the main channel and reference channel signals at the same time. However, the existing denoising methods are not effective in environments with low signal-to-noise ratios, and there is an urgent need for an effective sound denoising and enhancement method in complex environments. Summary of the invention
[0003] In view of the problems in the related art, the present invention proposes a method for noise reduction and speech enhancement with dual microphone cores to overcome the above-mentioned technical problems existing in the existing related art.
[0004] To this end, the specific technical solution adopted by the present invention is as follows:
[0005] In the first aspect, this patent provides a method for noise reduction and speech enhancement with dual microphone cores, comprising the following steps:
[0006] Step 1. Import two-channel sound signals, one of which is used as the main channel and the other as the reference channel;
[0007] Step 2. Use the reference channel signal and weights to estimate the noise through the minimum mean square error algorithm, calculate the error to continuously update the weight vector, and use the main channel signal to subtract the noise signal to obtain a single-channel sound signal;
[0008] Step 3. Perform Fourier transform on the signal, calculate the signal power spectrum density and noise power spectrum density, calculate the Wiener filter gain at each frequency point based on the calculation, apply the gain to the frequency domain signal of the current frame, and then perform inverse Fourier transform on the signal;
[0009] Step 4. Extract the harmonic signal, perform Fourier transform on the harmonic signal, use the harmonic regeneration coefficient to perform weighted average on the frequency domain amplitude of the enhanced signal and the frequency domain amplitude of the harmonic signal, and then perform inverse Fourier transform on the smoothed frequency domain signal.
[0010] In the second aspect, this patent provides a method with dual microphone core noise reduction and voice enhancement, including the following modules:
[0011] Microphone sampling module: includes two microphone cores for receiving dual-channel sound signals; the collected dual-channel sound signals are transmitted to the adaptive filtering module for processing;
[0012] Processor module: The adaptive filtering unit receives the dual-channel sound signal, performs the adaptive filtering process, and the processed signal is transmitted to the Wiener filtering unit; the Wiener filtering unit receives the single-channel sound signal after the adaptive filtering, performs the Wiener filtering process, and the processed signal is transmitted to the harmonic regeneration unit; the harmonic regeneration unit receives the single-channel sound signal after the Wiener filtering, performs the harmonic regeneration process, and the processed signal is transmitted to the output module;
[0013] Output module: Outputs the processed enhanced signal for further audio processing or direct playback through speakers.
[0014] Beneficial Effects of the Invention
[0015] The dual-channel sound signal is collected by a dual-microphone microphone, and the noise signal is estimated and extracted using an adaptive filtering method. After filtering out the noise, a single-channel sound signal is obtained. Then, the power spectral density of the signal and noise is calculated for each frame of the signal, and a Wiener filter is created and applied to obtain the denoised signal. Finally, the harmonic component is extracted, and smooth recursive processing is performed to generate a harmonic regeneration signal to obtain a denoised and enhanced output signal. By combining adaptive filtering, Wiener filtering and harmonic regeneration technology, it can effectively suppress noise and enhance voice signals in complex environments. The scope of application includes hearing aids, smart homes, speech recognition systems, remote conferencing systems and other scenarios that require high-quality audio input. The modular design of the system is flexible and scalable, and can be optimized according to different needs. BRIEF DESCRIPTION OF THE DRAWINGS
[0016] The accompanying drawings are used to provide a further understanding of the present invention and constitute a part of the specification. Together with the embodiments of the present invention, they are used to explain the present invention and do not constitute a limitation of the present invention. In the accompanying drawings:
[0017] Figure 1 Flowchart of the method for dual-microphone noise reduction and voice enhancement adopted in this patent.
[0018] Figure 2 A diagram showing the effect of a simulation using the denoising and enhancement method under high signal-to-noise ratio conditions.
[0019] Figure 3 A diagram showing the effect of a simulation using the denoising and enhancement method under low signal-to-noise ratio conditions.
[0020] Figure 4 Structural diagram of the method with dual microphone core noise reduction and voice enhancement adopted in this patent DETAILED DESCRIPTION
[0021] The following will be combined with the drawings in the embodiments of this patent to clearly and completely describe the technical solutions in the embodiments of this patent. Obviously, the described embodiments are only part of the embodiments of this patent, not all of them. Based on the embodiments in this patent, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this patent.
[0022] Figure 1 This is a flow chart of a dual-microphone-based speech enhancement method provided in an embodiment of the present invention. This embodiment mainly uses the dual-microphone-based speech enhancement method applied to a hearing aid device as an example to illustrate. The dual-microphone-based speech enhancement method provided in an embodiment of the present invention includes the following steps:
[0023] 101 is used to input two-channel sound signals, which are completed through a microphone core close to the sound source and another one slightly away from the sound source.
[0024] 102 is an adaptive filtering step, including:
[0025] The filter order is, the step factor is, the input signal from the reference channel is, and the filter coefficient is, then the output of the filter can be expressed as:
[0026] The desired signal is the sound signal from the main channel, and the error signal is the difference between the desired signal and the filter output, i.e., the filter coefficients are adjusted by minimizing the mean square error, and the update formula is:
[0027] The coefficients are continuously updated to obtain the output, which is an estimate of the noise. Therefore, to obtain the denoised noise, calculation is sufficient.
[0028] 103 is a Wiener filtering step, comprising:
[0029] The time domain signal is converted into the frequency domain signal through the fast Fourier transform (FFT), and its formula is expressed as:
[0030] Next, calculate the power spectral density of the signal and estimate the power spectral density of the noise using the following calculation method:
[0031] Among them, represents the noise power spectral density of the frame, and represents the corresponding forgetting factor. Therefore, the frequency domain transfer function of the Wiener filter can be calculated by adjusting the parameters:
[0032] Multiply the amplitude spectrum of the input signal by the Wiener filter gain to obtain the enhanced spectrum:
[0033] Finally, the frequency domain signal is converted to the time domain signal through inverse Fourier transform:
[0034] 104 is a harmonic regeneration step, comprising:
[0035] The negative part of the filtered signal is set to 0, and only the positive part is retained to extract the positive half of the signal, that is, to obtain the harmonic signal:
[0036] Through Fourier transform, the harmonic signal is converted from time domain to frequency domain:
[0037] Calculate the amplitude spectrum and phase spectrum of the harmonic signal using the frequency domain signal:
[0038] Taking the harmonic regeneration coefficient as the weighted average, the amplitude spectrum of the enhanced signal and the amplitude spectrum of the harmonic signal are smoothed and the spectrum is reconstructed:
[0039] Through inverse Fourier transform, the signal is converted from frequency domain to time domain again to obtain the regenerated harmonic signal:
[0040] After the signal is processed through the above steps, the output sound signal 105 is obtained.
[0041] In order to show that the method proposed in this patent has practical application value, Figure 2 and Figure 3 Two simulation effect diagrams are used for explanation and verification. Figure 2 This is a diagram of the denoising and enhancement effects achieved by the speech enhancement method described in this patent through computer simulation, and is based on an environment with a relatively high signal-to-noise ratio. Figure 3 This is a diagram of the denoising and enhancement effects achieved by the speech enhancement method described in this patent through computer simulation, and is based on an environment with a low signal-to-noise ratio.
[0042] exist Figure 2 In the figure, the first graphic window shows the clean signal (no noise) used for the simulation experiment, and the second and third graphic windows show the reception signals of the two microphone cores (corresponding to Figure 1 101 in ), the fourth graphic window shows the adaptive filtering step (corresponding to Figure 1 102 in the figure, the fifth graphic window shows the signal after the Wiener filtering step (corresponding to Figure 1 103 in the figure, the sixth graphic window shows the signal after the harmonic regeneration step (corresponding to Figure 1 The signal after 104). Figure 3 The meaning of each graphic window in Figure 2 Consistent.
[0043] exist Figure 2 and Figure 3 In the above, the distance between the two microphone cores is set to 45mm, the sound speed is set to 340m / s, and the adaptive filter order, step size factor, smoothing factor, and harmonic regeneration coefficient are set to 340m / s. The above values are obtained from practical experience, and this patent is not limited to this value method.
[0044] Figure 4 This is an embodiment of a method structure with dual-microphone core noise reduction and voice enhancement. The system can implement the method proposed in this patent, including a microphone sampling module 41, a processor module 42 and an output module 43.
[0045] In the microphone sampling module 41, 401 represents a microphone for collecting signals, which includes two microphone cores 402 and 403, 402 inputs the main channel signal, and 403 inputs the reference channel signal; in the processor module 42, 404 is capable of completing the adaptive filtering step (corresponding to Figure 1 102) of the processing unit, 405 is capable of completing the Wiener filtering step (corresponding to Figure 1 103) of the processing unit, 406 is capable of completing the harmonic regeneration step (corresponding to Figure 1 in the output module 43, the signal with denoising and enhancement effect is played through the speaker.
[0046] Those skilled in the art will appreciate that the embodiments of the present invention may be provided as methods, systems, or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware.
[0047] This patent is described with reference to the flowchart / block diagram of the method and system according to the embodiment of the patent. It should be understood that each process / block in the flowchart and / or block diagram, and the combination of the processes and / or blocks in the flowchart and / or block diagram can be implemented by computer program instructions.
[0048] The above are only embodiments of this patent and are not intended to limit this patent. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of this patent are included in the scope of the claims of this patent to be approved.
Claims
1. A method for noise reduction and speech enhancement with dual microphone cores, characterized in that: The following steps are involved: A two-channel sound signal is collected by a microphone with dual microphone cores; By using the adaptive filtering method, the noise signal is estimated and extracted, and the noise is filtered out to obtain a single-channel sound signal; Perform a fast Fourier transform on each frame of signal to convert the time domain signal to the frequency domain, then calculate the power spectral density of the signal and noise, create and apply the Wiener filter based on it, and then perform an inverse fast Fourier transform to convert the frequency domain signal to the time domain; Extract the harmonic components, perform fast Fourier transform on the harmonic signal, convert the time domain signal to the frequency domain, then perform smoothing recursive processing to generate a harmonic regeneration signal, and then perform an inverse fast Fourier transform to convert the frequency domain signal back to the time domain to obtain an outputtable denoised and enhanced signal.
2. The method for noise reduction and speech enhancement with dual microphone cores according to claim 1, characterized in that: Among the two channels of sound signals, one channel is the sound signal collected by the microphone core close to the sound source, called the main channel; the other channel is the sound signal collected by the microphone core slightly away from the sound source, called the reference channel.
3. The method for noise reduction and speech enhancement with dual microphone cores according to claim 1, characterized in that: The adaptive filtering method adopts the minimum mean square error algorithm, that is, the noise is estimated by the weight vector and the reference signal, and the error is calculated, and the filter weight vector is updated using the error signal and the input signal, so that the noise is extracted while minimizing the mean square error; The noise signal is subtracted from the main channel signal to obtain a denoised single-channel sound signal.
4. The method for noise reduction and speech enhancement with dual microphone cores according to claim 1, characterized in that: The Wiener filtering process first calculates the power spectral density of the signal, uses a forgetting factor to smooth the power spectral density of the noise, and updates the noise power spectral density estimate by comparing the power spectral density of the current frame with the noise power spectral density of the previous frame; According to the signal power spectrum density and the noise power spectrum density, the Wiener filter gain at each frequency point is calculated, and the calculated gain is used to adjust the amplitude of the frequency domain signal, retaining the phase of the original signal, and combining the adjusted amplitude and phase to obtain the filtered frequency domain signal; The filtered frequency domain signal is inverse Fourier transformed to obtain the time domain signal.
5. The method for noise reduction and speech enhancement with dual microphone cores according to claim 1, characterized in that: The harmonic regeneration process first sets the negative value part of the enhanced signal to zero, retaining only the positive value part to extract the harmonic component, and then performs a fast Fourier transform; The enhanced amplitude of the original signal and the amplitude of the harmonic signal are weighted averaged, where the weighting coefficient is the harmonic regeneration coefficient, which determines the degree of harmonic regeneration; The smoothed frequency domain signal is subjected to inverse Fourier transform to obtain the time domain signal again, which is the output signal that achieves denoising and enhancement effects.
6. A method for noise reduction and speech enhancement with dual microphone cores, characterized in that: It includes the following modules: microphone sampling module, adaptive filtering module, Wiener filtering module, harmonic regeneration module and output module.
7. The method for noise reduction and speech enhancement with dual microphone cores according to claim 6, characterized in that: The system can implement the sound denoising and enhancement method.
