A noise reduction method and system for a Bluetooth headset

Abstract

The present invention belongs to the field of noise reduction processing, and discloses a noise reduction method and system for a Bluetooth headset. The method includes: S1, obtaining a noise signal; S2, performing pre-emphasis processing on the noise signal to obtain a pre-emphasized signal; S3, performing frame division processing on the pre-emphasized signal to obtain a plurality of frame signals; S4, performing windowing processing on the frame signals to obtain windowed signals; S5, respectively performing noise reduction processing on each windowed signal to obtain a windowed signal after noise reduction; S6, performing feature extraction on the windowed signal after noise reduction to obtain signal features; S7, generating an anti-phase signal with the same signal features as the windowed signal after noise reduction but opposite in phase; S8, playing the anti-phase signals corresponding to each windowed signal after noise reduction in order of the acquisition time of the windowed signal after noise reduction from early to late. The present invention effectively reduces the occurrence probability of events where the pre-emphasis coefficient is too large or too small, improves the quality of the noise signal after pre-emphasis, and can more accurately obtain the features of the noise signal.

Description

Technical Field

[0001] The present invention relates to the field of noise reduction processing, and particularly to a noise reduction method and system for a Bluetooth headset. Background Art

[0002] The noise reduction methods of Bluetooth headsets generally include two types: active noise reduction and passive noise reduction. Active noise reduction captures external environmental noise signals by setting one or more microphones inside the headset, and then generates sound waves opposite to the noise signals (i.e., anti-phase signals) through the noise reduction circuit inside the headset, and cancels the noise by superposition, thereby achieving the noise reduction effect. Passive noise reduction mainly isolates external noise through the physical design and materials of the headset. This method usually relies on the sealing performance of the ear cups and the sound absorption characteristics of the materials.

[0003] During active noise reduction, it is necessary to first obtain the signal characteristics of the noise signal, and then generate anti-phase sound waves according to the signal characteristics, and play the anti-phase signal through the speaker of the headset to achieve active noise reduction. Before obtaining the characteristics of the noise signal, it is usually necessary to first perform pre-emphasis processing on the noise signal to obtain a pre-emphasized signal, and then perform subsequent operations such as frame segmentation, windowing, and signal feature extraction on the pre-emphasized signal. During the pre-emphasis processing, the prior art usually sets a fixed pre-emphasis coefficient. This makes the pre-emphasis processing unable to match the characteristics of the real-time noise signal, and it is easy to have the situation of too large or too small pre-emphasis coefficient. When the pre-emphasis coefficient is too large, the high-frequency signals in the noise signal will be over-enhanced, resulting in sound distortion, which is not conducive to subsequent signal feature extraction; when the pre-emphasis coefficient is too small, the high-frequency signals in the noise signal will be masked by the low-frequency signals, reducing the signal-to-noise ratio of the noise signal, which is also not conducive to subsequent signal feature extraction. Summary of the Invention

[0004] The purpose of the present invention is to disclose a noise reduction method and system for a Bluetooth headset, and solve the technical problem of how to more reasonably determine the pre-emphasis coefficient during the active noise reduction process of the Bluetooth headset, so as to improve the quality of the noise signal after pre-emphasis, so that the characteristics of the noise signal can be obtained more accurately, generate a more accurate anti-phase signal, and achieve a better active noise reduction effect.

[0005] In order to achieve the above purpose, the present invention adopts the following technical solutions:

[0006] On the one hand, the present invention provides a noise reduction method for a Bluetooth headset, including:

[0007] S1, obtaining a noise signal;

[0008] S2, performing pre-emphasis processing on the noise signal to obtain a pre-emphasized signal;

[0009] S3. Perform frame processing on the pre-emphasized signal to obtain multiple frame signals;

[0010] S4. Perform windowing on the frame signals to obtain windowed signals;

[0011] S5. Perform noise reduction on each windowed signal respectively to obtain the windowed signals after noise reduction;

[0012] S6. Extract features from the windowed signals after noise reduction to obtain signal features;

[0013] S7. Generate an anti-phase signal with the same signal features as the windowed signals after noise reduction and opposite in phase;

[0014] S8. Play the anti-phase signals corresponding to each windowed signal after noise reduction in the order of the acquisition time of the windowed signals after noise reduction from early to late;

[0015] Among them, performing pre-emphasis processing on the noise signal to obtain the pre-emphasized signal includes:

[0016] S10. Perform the following processing on each sampling point in the noise signal respectively:

[0017] When q is less than the set threshold qthre, perform pre-emphasis processing on the qth sampling point x(q) in the noise signal using a preset pre-emphasis coefficient;

[0018] When q is greater than or equal to the set threshold qthre, for the qth sampling point x(q) in the noise signal, calculate its corresponding pre-emphasis coefficient p based on the frequency characteristics of the neighborhood signal of the sampling point x(q) q :

[0019] Use the pre-emphasis coefficient p q Perform pre-emphasis processing on x(q) to obtain the pre-emphasized sampling point y(q);

[0020] S20. The pre-emphasized signal is composed of all the pre-emphasized sampling points obtained by S10.

[0021] Preferably, S1 includes:

[0022] Obtain the analog signal of the sound in the environment where the earphone is located;

[0023] Perform discrete sampling on the analog signal to obtain the noise signal.

[0024] Preferably, performing windowing on the frame signals to obtain the windowed signals includes:

[0025] Perform windowing on each frame signal respectively using a preset window function to obtain the windowed signals.

[0026] Preferably, the preset window function includes any one of a Hanning window, a Hamming window, and a Blackman window.

[0027] Preferably, the pre-emphasized signal is framed to obtain a plurality of frame signals, including:

[0028] Let T represent the preset frame length;

[0029] Use T to perform non-overlapping framing on the pre-emphasized signal to obtain a plurality of temporary frames;

[0030] Calculate the overlapping ratio based on the temporary frames;

[0031] Perform overlapping framing on the pre-emphasized signal based on the overlapping ratio to obtain a plurality of frame signals.

[0032] Preferably, use T to perform non-overlapping framing on the pre-emphasized signal to obtain a plurality of temporary frames, including:

[0033] For the k-th temporary frame, its corresponding time interval is len represents the total duration of the pre-emphasized signal.

[0034] Preferably, calculating the overlapping ratio based on the temporary frames includes:

[0035] Calculate the fluctuation coefficients between two adjacent temporary frames respectively:

[0036] For the k-th and the (k + 1)-th adjacent temporary frames tempsig k and tempsig k+1 , calculate the fluctuation coefficient volcoef k between tempsig k+1 and tempsig k,k+1 ;

[0037] Use the following formula to calculate the overlapping ratio:

[0038]

[0039] overat represents the overlapping ratio, ratmi represents the preset ratio value, len represents the total duration of the pre-emphasized signal, volceof ave represents the average value of mak fluctuation coefficients, volceof mid represents the median value of mak fluctuation coefficients, volceof max represents the maximum value of mak fluctuation coefficients; α1 and α2 respectively represent the first fluctuation weight and the second fluctuation weight; Q represents the preset positive integer, and β1 and β2 respectively represent the fixed ratio coefficient and the variable ratio coefficient.

[0040] Preferably, the fluctuation coefficient volcoef k,k+1 is calculated as follows:

[0041] Let t str,k and t end,k represent the start time and end time of the k-th temporary frame tempsig k respectively; let t str,k+1 and t end,k+1 represent the start time and end time of the (k + 1)-th temporary frame tempsig k+1 respectively;

[0042] Obtain the set sampU of sampling points in the pre-emphasized signal within the time interval ;

[0043] Calculate volcoef k,k+1 using the following formula:

[0044]

[0045] NsampU represents the total number of sampling points in sampU, and range i represents the amplitude of the i-th sampling point.

[0046] Preferably, calculate the corresponding pre-emphasis coefficient p q based on the neighborhood signal frequency characteristics of the sampling point x(q), including:

[0047] Calculate p q using the following formula:

[0048]

[0049] p rst represents the base value of the pre-emphasis coefficient, f q represents the estimated frequency of the signal within the time interval in the noise signal, t q represents the sampling time corresponding to the q-th sampling point, H represents the sampling interval, f q-1 represents the estimated frequency of the signal within the time interval in the noise signal, t q-1 represents the sampling time corresponding to the (q + 1)-th sampling point, max represents obtaining the larger value between f q and f q-1 , η1 represents the first weighting coefficient, and η2 represents the second weighting coefficient.

[0050] On the other hand, the present invention provides a noise reduction system for a Bluetooth headset, including a sound collection module, a pre-emphasis module, a windowing module, a framing module, a noise reduction module, a feature extraction module, a generation module, and a playback module;

[0051] The sound collection module is used to obtain noise signals;

[0052] The pre-emphasis module is used to perform pre-emphasis processing on the noise signals to obtain pre-emphasized signals;

[0053] The windowing module is used to perform windowing processing on the frame signals to obtain windowed signals;

[0054] The frame segmentation module is used to perform frame segmentation processing on the pre-emphasized signals to obtain multiple frame signals;

[0055] The noise reduction module is used to perform noise reduction processing on each of the frame signals to obtain the frame signals after noise reduction;

[0056] The feature extraction module is used to perform feature extraction on the frame signals after noise reduction to obtain signal features;

[0057] The generation module is used to generate anti-phase signals that have the same signal features as the frame signals after noise reduction but opposite phases;

[0058] The playback module is used to sequentially play the anti-phase signals corresponding to each of the frame signals after noise reduction in the order of the acquisition time of the frame signals after noise reduction from early to late;

[0059] Among them, performing pre-emphasis processing on the noise signals to obtain pre-emphasized signals includes:

[0060] S10, performing the following processing on each sampling point in the noise signals respectively:

[0061] When q is less than the set threshold qthre, perform pre-emphasis processing on the qth sampling point x(q) in the noise signals using a preset pre-emphasis coefficient;

[0062] When q is greater than or equal to the set threshold qthre, for the qth sampling point x(q) in the noise signals, calculate its corresponding pre-emphasis coefficient p based on the frequency characteristics of the neighborhood signals of the sampling point x(q) q :

[0063] Use the pre-emphasis coefficient p q to perform pre-emphasis processing on x(q) to obtain the pre-emphasized sampling point y(q);

[0064] S20, the pre-emphasized signals are composed of all the pre-emphasized sampling points obtained from S10.

[0065] Beneficial effects:

[0066] Compared with existing active noise reduction methods, in the process of pre-emphasizing the noise signal, the present invention calculates the pre-emphasis coefficient corresponding to the sampling point based on the frequency characteristics of the neighborhood signal of the sampling point, and then pre-emphasizes the noise signal based on the calculated pre-emphasis coefficient, so that the pre-emphasis coefficient can adaptively change with the change of the neighborhood signal frequency characteristics, effectively reducing the occurrence probability of events where the pre-emphasis coefficient is too large or too small, thus a more reasonable pre-emphasis coefficient can be calculated, effectively improving the quality of the pre-emphasized noise signal, which is beneficial to obtaining the characteristics of the noise signal more accurately, generating a more accurate anti-phase signal, and achieving a better active noise reduction effect. BRIEF DESCRIPTION OF THE DRAWINGS

[0067] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following will briefly introduce the drawings required in the embodiments. It should be understood that the following drawings only show some embodiments of the present invention, and therefore should not be regarded as limiting the scope. For those of ordinary skill in the art, without creative efforts, other related drawings can also be obtained based on these drawings.

[0068] Figure 1 It is a schematic diagram of a noise reduction method for a Bluetooth headset according to the present invention.

[0069] Figure 2 It is a schematic diagram of a noise reduction system for a Bluetooth headset according to the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0070] The following will clearly and completely describe the technical solutions in the embodiments of the present invention with reference to the drawings in the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, rather than all embodiments. The components of the embodiments of the present invention described and illustrated herein can be arranged and designed in various different configurations. Therefore, the detailed description of the embodiments of the present invention provided in the following drawings is not intended to limit the scope of the claimed invention, but merely represents selected embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts belong to the scope of protection of the present invention.

[0071] As Figure 1 shown in an embodiment, the present invention provides a noise reduction method for a Bluetooth headset, including:

[0072] S1, obtaining a noise signal;

[0073] S2, performing pre-emphasis processing on the noise signal to obtain a pre-emphasized signal;

[0074] S3, performing frame division processing on the pre-emphasized signal to obtain a plurality of frame signals;

[0075] S4. Window the frame signal to obtain a windowed signal;

[0076] S5. Denoise each windowed signal to obtain a denoised windowed signal;

[0077] S6. Extract features from the denoised windowed signal to obtain signal features;

[0078] S7. Generate an anti-phase signal that has the same signal features as the denoised windowed signal but with the opposite phase;

[0079] S8. Play the anti-phase signal corresponding to each denoised windowed signal in order of the acquisition time of the denoised windowed signal from earliest to latest;

[0080] Among them, pre-emphasize the noise signal to obtain a pre-emphasized signal, including:

[0081] S10. Process each sampling point in the noise signal as follows:

[0082] When q is less than the set threshold qthre, pre-emphasize the q-th sampling point x(q) in the noise signal using a preset pre-emphasis coefficient;

[0083] When q is greater than or equal to the set threshold qthre, for the q-th sampling point x(q) in the noise signal, calculate its corresponding pre-emphasis coefficient p based on the frequency characteristics of the neighborhood signal of the sampling point x(q) q :

[0084] Use the pre-emphasis coefficient p q To pre-emphasize x(q) to obtain a pre-emphasized sampling point y(q);

[0085] S20. The pre-emphasized signal is composed of all the pre-emphasized sampling points obtained from S10.

[0086] During the pre-emphasis of the noise signal, calculate the pre-emphasis coefficient corresponding to the sampling point based on the frequency characteristics of the neighborhood signal of the sampling point, and then pre-emphasize the noise signal based on the calculated pre-emphasis coefficient, so that the pre-emphasis coefficient can adaptively change with the change of the neighborhood signal frequency characteristics, effectively reducing the occurrence probability of events where the pre-emphasis coefficient is too large or too small. Thus, a more reasonable pre-emphasis coefficient can be calculated, effectively improving the quality of the pre-emphasized noise signal, which is beneficial to obtaining the characteristics of the noise signal more accurately, generating a more accurate anti-phase signal, and achieving a better active noise reduction effect.

[0087] Pre-emphasis is a commonly used technique in signal processing. Its main purpose is to improve the characteristics of a signal by boosting the energy of high-frequency components. Pre-emphasis increases the energy of high-frequency components, ensuring that these components are not drowned out or ignored due to low energy during subsequent processing (such as transmission, encoding, etc.). This makes them more prominent in signal transmission and helps improve the signal-to-noise ratio, especially for the high-frequency part. Since noise is usually relatively evenly distributed across the entire spectrum.

[0088] Preferably, the set threshold qthre is 40.

[0089] Preferably, S1 includes:

[0090] Obtain an analog signal of the sound in the environment where the earphone is located;

[0091] Perform discrete sampling on the analog signal to obtain a noise signal.

[0092] Specifically, the analog signal is discretely sampled at a fixed sampling interval to obtain a noise signal.

[0093] Preferably, obtaining an analog signal of the sound in the environment where the earphone is located includes:

[0094] Continuously obtain an analog signal of the sound in the environment where the earphone is located, and clip the obtained analog signal according to a set length to obtain a noise signal.

[0095] Specifically, after the user activates the active noise reduction function of the present invention, the acquisition of the noise signal in the environment is continuously carried out, and noise signals that are temporally connected and have a length equal to the set length are continuously obtained.

[0096] The noise reduction module in the Bluetooth earphone continuously generates an anti-phase signal based on the continuously obtained signal and plays it through the speaker in the Bluetooth earphone to achieve the effect of canceling out noise.

[0097] Preferably, the set length is 0.2 seconds.

[0098] Preferably, obtaining an analog signal of the sound in the environment where the earphone is located includes:

[0099] Use a microphone set outside the earphone to obtain an analog signal of the sound in the environment where the earphone is located.

[0100] Preferably, windowing the frame signal to obtain a windowed signal includes:

[0101] Respectively use a preset window function to window each frame signal to obtain a windowed signal.

[0102] In the process of signal processing, especially in frequency domain analysis, windowing is a commonly used technique to reduce the impact of truncation effects (i.e., spectral leakage). Since signals are usually non-periodic, this truncation can lead to spectral leakage, affecting the accuracy of the spectrum. Windowing can make the truncation boundary smoother, thereby reducing the impact of spectral leakage. Different window functions have different abilities to suppress spectral leakage. Different window functions have different main lobe widths and sidelobe suppression capabilities. Selecting an appropriate window function can achieve a balance between frequency resolution and sidelobe suppression. For example, the Hanning window and Hamming window can effectively reduce sidelobes but slightly increase the main lobe width, which is suitable for general signal processing; while the flat-top window can better measure the amplitude of the signal, but has a larger main lobe width and is suitable for applications that require precise amplitude measurement. When the signal is processed in frames, if it is directly truncated without windowing, it will lead to discontinuities at the frame edges, thereby introducing high-frequency components and other unwanted effects. Windowing can make the signal gradually become zero at the edges, thus reducing the impact of edge effects.

[0103] Preferably, the preset window function includes any one of the Hanning window, Hamming window, and Blackman window.

[0104] Preferably, the pre-emphasized signal is processed in frames to obtain a plurality of frame signals, including:

[0105] Let T represent the preset frame length;

[0106] The pre-emphasized signal is non-overlappingly framed using T to obtain a plurality of temporary frames;

[0107] Calculate the overlap ratio based on the temporary frames;

[0108] The pre-emphasized signal is overlappingly framed based on the overlap ratio to obtain a plurality of frame signals.

[0109] In the prior art, a fixed overlap ratio is generally used. Obviously, in this framing method, since the actual changes of the signal are not taken into account, problems such as too high or too low overlap ratio are likely to occur. If the overlap ratio is too small, that is, the frame shift is too large, it is easy to cause problems such as signal information loss, insufficient smoothness, and inaccurate spectrum analysis, which will affect the accuracy of the anti-phase signal generated subsequently in the present invention. And if the overlap ratio is too high, that is, the frame shift is too small, it will lead to data redundancy, too low computational efficiency, and loss of sensitivity to signal details.

[0110] Therefore, the present invention calculates the overlapping ratio through temporary frames, which can make the overlapping ratio adapt to the change of the signal, so as to obtain a better frame segmentation result. While ensuring the quality of the frame segmentation result, it can effectively control the number of frames obtained after frame segmentation to avoid affecting the calculation efficiency of subsequent calculations.

[0111] Preferably, the value range of the preset frame length is [20ms, 40ms].

[0112] As a further preference, the value of T is 30ms.

[0113] Preferably, the pre-emphasized signal is non-overlappingly framed using T to obtain a plurality of temporary frames, including:

[0114] For the k-th temporary frame, its corresponding time interval is len represents the total duration of the pre-emphasized signal.

[0115] After obtaining the temporary frames through non-overlapping framing, the overall change of the signal in the overlapping part during overlapping framing can be effectively estimated, so that a more accurate overlapping ratio can be obtained based on the change.

[0116] Preferably, calculating the overlapping ratio based on the temporary frames includes:

[0117] Calculating the fluctuation coefficients between two adjacent temporary frames respectively:

[0118] For the k-th and the (k + 1)-th adjacent temporary frames tempsig k and tempsig k+1 , calculate the fluctuation coefficient volcoef k between tempsig k+1 ; k,k+1 ;

[0119] Calculating the overlapping ratio using the following formula:

[0120]

[0121] overat represents the overlapping ratio, ratmi represents the preset ratio value, len represents the total duration of the pre-emphasized signal, volceof ave represents the average value of mak fluctuation coefficients, volceof mid represents the median value of mak fluctuation coefficients, volceof max represents the maximum value of mak fluctuation coefficients; α1 and α2 respectively represent the first fluctuation weight and the second fluctuation weight; Q represents the preset positive integer, and β1 and β2 respectively represent the fixed ratio coefficient and the variable ratio coefficient.

[0122] The overlapping ratio of the present invention is calculated based on the fluctuation coefficients between all pairs of temporary frames. The more drastic and larger the change in the fluctuation coefficient, and the larger the median value, the higher the overlapping ratio. Thus, when the change in noise is more drastic, a larger overlapping ratio can be adopted to ensure the smooth transition of the signal after frame division, ensure that the change in the signal within each frame is small, and thus make the feature extraction and analysis more accurate. Since the present invention uses the same frame length as that of the subsequent overlapping frame division when obtaining temporary frames in the non-overlapping frame division manner, the calculated fluctuation coefficient can more accurately represent the change situation of the overlapping part between two adjacent frames after overlapping frame division, improving the accuracy of the calculated overlapping ratio.

[0123] Preferably, the preset proportional value is 0.4.

[0124] Preferably, the first fluctuation weight and the second fluctuation weight are 0.7 and 0.3 respectively.

[0125] Preferably, the preset positive integer is 2.

[0126] Preferably, the fixed proportional coefficient and the variable proportional coefficient are 0.6 and 0.4 respectively.

[0127] The first fluctuation weight, the second fluctuation weight, the fixed proportional coefficient, and the variable proportional coefficient can all be adjusted according to the actual situation, and the above are some preferred embodiments.

[0128] Preferably, the fluctuation coefficient volcoef k,k+1 is calculated as follows:

[0129] Use t str,k and t end,k to represent the start time and end time of the k-th temporary frame tempsig k respectively; use t str,k+1 and t end,k+1 to represent the start time and end time of the (k + 1)-th temporary frame tempsig k+1 respectively;

[0130] Obtain the set sampU of the sampling points in the pre-emphasized signal within the time interval ;

[0131] Calculate volcoef k,k+1 using the following formula:

[0132]

[0133] NsampU represents the total number of sampling points in sampU, and range i represents the amplitude of the sampling point i.

[0134] The fluctuation coefficient of the present invention calculates the fluctuation of the signal in the interval composed of the second half of the previous temporary frame and the first half of the subsequent temporary frame among two temporary frames. By calculating the standard deviation of the amplitude, the fluctuation of the amplitude can be effectively obtained.

[0135] Preferably, overlapping framing processing is performed on the pre-emphasized signal based on the overlapping ratio to obtain multiple frame signals, including:

[0136] The pre-emphasized signal is framed using T. The length of the same signal included between two adjacent frame signals divided by T is equal to the overlapping ratio, that is, frame shift = T×(1 - overat), where overat represents the overlapping ratio.

[0137] Specifically, the higher the overlapping ratio, the smaller the frame shift.

[0138] Preferably, the q-th sampling point x(q) in the noise signal is pre-emphasized using a preset pre-emphasis coefficient, including:

[0139] Using p pre to represent the preset pre-emphasis coefficient, when pre-emphasizing x(q), the processing formula is as follows:

[0140] y(q) = x(q) - p pre ×x(q - 1)

[0141] y(q) represents the sampling point obtained after pre-emphasizing x(q), and x(q - 1) represents the (q - 1)-th sampling point.

[0142] Preferably, the value of p pre is 0.95.

[0143] Preferably, the corresponding pre-emphasis coefficient p is calculated based on the frequency characteristics of the neighborhood signal of the sampling point x(q) q , including:

[0144] The following formula is used to calculate p q :

[0145]

[0146] p rst represents the base value of the pre-emphasis coefficient, f q represents the estimated frequency of the signal in the noise signal within the time interval , t q represents the sampling moment corresponding to the q-th sampling point, H represents the sampling interval, f q-1 represents the estimated frequency of the signal in the noise signal within the time interval t q-1denotes the sampling moment corresponding to the (q + 1)-th sampling point, max denotes obtaining the larger value between f q and f q-1 , η1 denotes the first weighting coefficient, η2 denotes the second weighting coefficient, and S denotes the interval control parameter.

[0147] The pre-emphasis coefficient of the present invention is calculated based on the change in the estimated frequencies of the signals in two time intervals adjacent to the current sampling point. When the estimated frequency shows an increasing trend, the calculated pre-emphasis coefficient of the present invention will also be amplified accordingly. Therefore, when the estimated frequency of the signal in the time interval is higher and the increase amplitude of the estimated frequencies of the signals in two adjacent time intervals is larger, the pre-emphasis coefficient is larger, thereby effectively enhancing the high-frequency part in the noise to obtain more accurate signal features when performing feature extraction subsequently. The pre-emphasis coefficient calculated by the present invention can change with the change trend of the estimated frequency of the signal before the sampling point. Therefore, it can more effectively enhance the high-frequency part in the noise, and at the same time, it can also reduce the occurrence probability of events such as over-enhancement or signal-to-noise ratio reduction caused by an excessive fixed pre-emphasis coefficient, improving the pre-emphasis effect.

[0148] Preferably, the base value of the pre-emphasis coefficient is 0.9.

[0149] Preferably, the sampling interval is the reciprocal of the sampling rate.

[0150] Preferably, the sampling rate is 30 kHz.

[0151] Preferably, the first weighting coefficient and the second weighting coefficient are 0.8 and 0.2 respectively.

[0152] Preferably,

[0153] Preferably, in the process of estimating the frequency of the signal, the zero-crossing rate method is used to obtain the estimated frequency of the signal by calculating the number of times the signal crosses zero.

[0154] The LPC algorithm can also be used to estimate the frequency of the signal.

[0155] Preferably, using the pre-emphasis coefficient p q to perform pre-emphasis processing on x(q) to obtain the pre-emphasized sampling point y(q), including:

[0156] y(q) = x(q) - p q ×x(q - 1)

[0157] x(q - 1) represents the (q - 1)-th sampling point.

[0158] Preferably, noise reduction processing is respectively performed on each windowed signal to obtain a windowed signal after noise reduction, including:

[0159] Calculating the selection coefficient of the windowed signal;

[0160] If the selection coefficient is greater than the set selection coefficient threshold, the first noise reduction algorithm is used to perform noise reduction on the windowed signal;

[0161] If the selection coefficient is less than or equal to the set selection coefficient threshold, the second noise reduction algorithm is used to perform noise reduction on the windowed signal.

[0162] The present invention selects the noise reduction algorithm based on the selection coefficient, and can obtain a more appropriate noise reduction algorithm based on the signal characteristics of the windowed signal, improving the noise reduction effect.

[0163] Preferably, for the z-th windowed signal Fra z , the calculation formula of its corresponding selection coefficient is as follows:

[0164] chosv z =|vras z-1 -vras z |

[0165] chosv z represents the selection coefficient corresponding to Fra z , vras z and vras z-1 respectively represent the change parameters corresponding to Fra z and Fra z-1 , Fra z-1 represents the (z - 1)-th windowed signal, and the calculation formula of vras z is:

[0166]

[0167] sguz represents the set of sampling points included in Fra z , range v represents the amplitude of the sampling point v, and nsguz represents the total number of sampling points in sguz;

[0168] vras z-1 is calculated in the same way as vras z .

[0169] The selection coefficient compares the difference between the average values of the amplitudes of the sampling points between two windowed signals. If the selection coefficient is larger, it means the change amplitude between the two windowed signals is larger.

[0170] Preferably, the selection coefficient threshold is 0.05.

[0171] Preferably, the first noise reduction algorithm includes a wavelet transform noise reduction algorithm and a noise reduction algorithm based on a convolutional neural network.

[0172] Preferably, the second noise reduction algorithm includes a noise reduction algorithm based on spectral subtraction.

[0173] When the change amplitude between the two windowed signals is larger, the probability that the present invention selects the first noise reduction algorithm with better noise reduction effect for non-stationary signals is greater, while when the change amplitude is smaller, the probability of selecting the second noise reduction algorithm with better noise reduction effect for stationary signals is greater, realizing the adaptive selection of the noise reduction algorithm.

[0174] Preferably, the signal features include amplitude and frequency.

[0175] Preferably, generating an anti-phase signal with the same signal features as the noise-reduced windowed signal but opposite in phase includes:

[0176] Generating an anti-phase signal with the same amplitude, frequency, and length as the noise-reduced windowed signal but opposite in phase.

[0177] As Figure 2 shown in an embodiment, the present invention provides a noise reduction system for a Bluetooth headset, including a sound collection module, a pre-emphasis module, a windowing module, a framing module, a noise reduction module, a feature extraction module, a generation module, and a playback module;

[0178] The sound collection module is used to acquire a noise signal;

[0179] The pre-emphasis module is used to perform pre-emphasis processing on the noise signal to obtain a pre-emphasized signal;

[0180] The windowing module is used to perform windowing processing on the framed signal to obtain a windowed signal;

[0181] The framing module is used to perform framing processing on the pre-emphasized signal to obtain a plurality of framed signals;

[0182] The noise reduction module is used to perform noise reduction processing on each framed signal respectively to obtain a noise-reduced framed signal;

[0183] The feature extraction module is used to extract features from the noise-reduced framed signal to obtain signal features;

[0184] The generation module is used to generate an anti-phase signal with the same signal features as the noise-reduced framed signal but opposite in phase;

[0185] The playback module is used to play the anti-phase signals corresponding to each noise-reduced framed signal in order of the acquisition time of the noise-reduced framed signals from earliest to latest;

[0186] Among them, performing pre-emphasis processing on the noise signal to obtain a pre-emphasized signal includes:

[0187] S10. Process each sampling point in the noise signal as follows:

[0188] When q is less than the set threshold qthre, perform pre-emphasis processing on the qth sampling point x(q) in the noise signal using a preset pre-emphasis coefficient;

[0189] When q is greater than or equal to the set threshold qthre, for the qth sampling point x(q) in the noise signal, calculate its corresponding pre-emphasis coefficient p based on the frequency characteristics of the neighborhood signal of the sampling point x(q) q :

[0190] Use the pre-emphasis coefficient p q to perform pre-emphasis processing on x(q) to obtain the pre-emphasized sampling point y(q);

[0191] S20. The pre-emphasized signal is composed of all the pre-emphasized sampling points obtained in S10.

[0192] As described above, it is only the specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art within the technical scope disclosed by the present invention can easily think of changes or substitutions, which should all be covered within the protection scope of the present invention. Therefore, the protection scope of the present invention should be subject to the protection scope of the claims.

Claims

1. A Bluetooth headset noise reduction method, characterized in that: include: S1, obtain noise signal; S2, pre-emphasize the noise signal to obtain a pre-emphasized signal; S3, performing frame processing on the pre-emphasized signal to obtain multiple frame signals; S4, performing windowing processing on the frame signal to obtain a windowed signal; S5, performing noise reduction processing on each windowed signal to obtain a windowed signal after noise reduction; S6, extracting features from the windowed signal after noise reduction to obtain signal features; S7, generating an anti-phase signal having the same signal characteristics as the windowed signal after noise reduction and opposite phase; S8, playing the inverted signal corresponding to each noise-reduced windowed signal in sequence from early to late according to the acquisition time of the noise-reduced windowed signal; The pre-emphasis processing is performed on the noise signal to obtain the pre-emphasis signal, including: S10, performing the following processing on each sampling point in the noise signal: When q is less than the set threshold qthre, a preset pre-emphasis coefficient is used to perform pre-emphasis processing on the qth sampling point x(q) in the noise signal; When q is greater than or equal to the set threshold qthre, for the qth sampling point x(q) in the noise signal, the corresponding pre-emphasis coefficient p is calculated based on the frequency characteristics of the neighborhood signal of the sampling point x(q) q : Use the pre-emphasis factor p q Perform pre-emphasis processing on x(q) to obtain the pre-emphasized sampling point y(q); S20, all the pre-emphasized sampling points obtained in S10 form a pre-emphasized signal.

2. A Bluetooth headset noise reduction method according to claim 1, characterized in that: S1 includes: Acquire an analog signal of the sound of the environment in which the earphone is located; The analog signal is discretely sampled to obtain a noise signal.

3. A Bluetooth headset noise reduction method according to claim 1, characterized in that: Perform windowing processing on the frame signal to obtain a windowed signal, including: Use a preset window function to perform windowing processing on each frame signal to obtain a windowed signal.

4. A Bluetooth headset noise reduction method according to claim 3, characterized in that: The preset window functions include any one of the Hanning window, the Hamming window and the Leckman window.

5. A Bluetooth headset noise reduction method according to claim 1, characterized in that: The pre-emphasized signal is framed to obtain multiple frame signals, including: T represents the preset frame length; Use T to perform non-overlapping framing on the pre-emphasized signal to obtain multiple temporary frames; Calculate the overlap ratio based on the temporary frame; The pre-emphasized signal is subjected to overlapping frame division processing based on the overlapping ratio to obtain a plurality of frame signals.

6. A Bluetooth headset noise reduction method according to claim 5, characterized in that: Use T to perform non-overlapping framing on the pre-emphasized signal to obtain multiple temporary frames, including: For the kth temporary frame, the corresponding time interval is len represents the total duration of the pre-emphasis signal.

7. A Bluetooth headset noise reduction method according to claim 5, characterized in that: Calculate the overlap ratio based on the temporary frame, including: Calculate the fluctuation coefficient between two adjacent temporary frames respectively: For the kth and k+1th adjacent temporary frames tempsig k and tempsig k+1 , calculate tempsig k and tempsig k+1 The coefficient of fluctuation between k,k+1 ; The overlap ratio is calculated using the following formula: Overat indicates the overlap ratio, ratmi indicates the preset ratio value, len represents the total duration of the pre-emphasized signal, volceof ave Represents the average value of mak volatility coefficients, volceof mid Represents the median of mak volatility coefficients, volceof max Represents the maximum value of mak volatility coefficients; α1 and α2 represent the first volatility weight and the second volatility weight respectively; Q represents a preset positive integer, β1 and β2 represent the fixed proportional coefficient and the variable proportional coefficient respectively.

8. A Bluetooth headset noise reduction method according to claim 7, characterized in that: Volcoef k,k+1 The calculation process is as follows: Use t str,k and t end,k Indicates the kth temporary frame tempsig k The start and end time of str,k+1 and t end,k+1 Indicates the k+1th temporary frame tempsig k+1 The start and end time of Get the time interval of the pre-emphasized signal The set of sampling points sampU in ; Calculate volcoef using the following formula k,k+1 : NsampU represents the total number of sampling points in sampU, range i Represents the amplitude of sampling point i.

9. A Bluetooth headset noise reduction method according to claim 1, characterized in that: Calculate the corresponding pre-emphasis coefficient p based on the frequency characteristics of the neighborhood signal of the sampling point x(q) q ,include: Use the following formula to calculate p q : p rst Indicates the base value of the pre-emphasis coefficient, f q Indicates that in the noise signal, in the time interval The estimated frequency of the signal in , t q represents the sampling time corresponding to the qth sampling point, H represents the sampling interval, and f q-1 Indicates that in the noise signal, in the time interval The estimated frequency of the signal in , t q-1 Indicates the sampling time corresponding to the q+1th sampling point, and max indicates the acquisition of f q and f q-1 The larger value between η1 and η2, η1 represents the first weighting coefficient, η2 represents the second weighting coefficient; S represents the interval control parameter.

10. A Bluetooth headset noise reduction system, characterized in that: It includes a sound receiving module, a pre-emphasis module, a windowing module, a framing module, a noise reduction module, a feature extraction module, a generation module and a playback module; The radio module is used to obtain noise signals; The pre-emphasis module is used to perform pre-emphasis processing on the noise signal to obtain a pre-emphasis signal; The windowing module is used to perform windowing processing on the frame signal to obtain a windowed signal; The framing module is used to perform framing processing on the pre-emphasized signal to obtain multiple frame signals; The noise reduction module is used to perform noise reduction processing on each frame signal to obtain a noise-reduced frame signal; The feature extraction module is used to extract features from the frame signal after noise reduction to obtain signal features; The generating module is used to generate an anti-phase signal having the same signal characteristics as the frame signal after noise reduction and opposite phase; The playback module is used to play the inverted signal corresponding to each noise-reduced frame signal in sequence from early to late according to the acquisition time of the noise-reduced frame signal; The pre-emphasis processing is performed on the noise signal to obtain the pre-emphasis signal, including: S10, performing the following processing on each sampling point in the noise signal: When q is less than the set threshold qthre, a preset pre-emphasis coefficient is used to perform pre-emphasis processing on the qth sampling point x(q) in the noise signal; When q is greater than or equal to the set threshold qthre, for the qth sampling point x(q) in the noise signal, the corresponding pre-emphasis coefficient p is calculated based on the frequency characteristics of the neighborhood signal of the sampling point x(q) q : Use the pre-emphasis factor p q Perform pre-emphasis processing on x(q) to obtain the pre-emphasized sampling point y(q); S20, all the pre-emphasized sampling points obtained in S10 form a pre-emphasized signal.

Images