WO2020114071A1 - 用于采集体音信号的双麦克风自适应滤波算法及应用 - Google Patents

用于采集体音信号的双麦克风自适应滤波算法及应用 Download PDF

Info

Publication number
WO2020114071A1
WO2020114071A1 PCT/CN2019/110287 CN2019110287W WO2020114071A1 WO 2020114071 A1 WO2020114071 A1 WO 2020114071A1 CN 2019110287 W CN2019110287 W CN 2019110287W WO 2020114071 A1 WO2020114071 A1 WO 2020114071A1
Authority
WO
WIPO (PCT)
Prior art keywords
signal
microphone
body sound
pass filtering
adaptive
Prior art date
Application number
PCT/CN2019/110287
Other languages
English (en)
French (fr)
Inventor
莫鸿强
田翔
Original Assignee
华南理工大学
佛山市百步梯医疗科技有限公司
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 华南理工大学, 佛山市百步梯医疗科技有限公司 filed Critical 华南理工大学
Priority to SG11202103089YA priority Critical patent/SG11202103089YA/en
Priority to US17/282,761 priority patent/US11735200B2/en
Publication of WO2020114071A1 publication Critical patent/WO2020114071A1/zh

Links

Images

Classifications

    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B7/00—Instruments for auscultation
    • A61B7/02—Stethoscopes
    • A61B7/026—Stethoscopes comprising more than one sound collector
    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B7/00—Instruments for auscultation
    • A61B7/02—Stethoscopes
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04R—LOUDSPEAKERS, MICROPHONES, GRAMOPHONE PICK-UPS OR LIKE ACOUSTIC ELECTROMECHANICAL TRANSDUCERS; DEAF-AID SETS; PUBLIC ADDRESS SYSTEMS
    • H04R3/00—Circuits for transducers, loudspeakers or microphones
    • H04R3/005—Circuits for transducers, loudspeakers or microphones for combining the signals of two or more microphones
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04R—LOUDSPEAKERS, MICROPHONES, GRAMOPHONE PICK-UPS OR LIKE ACOUSTIC ELECTROMECHANICAL TRANSDUCERS; DEAF-AID SETS; PUBLIC ADDRESS SYSTEMS
    • H04R3/00—Circuits for transducers, loudspeakers or microphones
    • H04R3/04—Circuits for transducers, loudspeakers or microphones for correcting frequency response
    • G—PHYSICS
    • G10—MUSICAL INSTRUMENTS; ACOUSTICS
    • G10L—SPEECH ANALYSIS TECHNIQUES OR SPEECH SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING TECHNIQUES; SPEECH OR AUDIO CODING OR DECODING
    • G10L21/00—Speech or voice signal processing techniques to produce another audible or non-audible signal, e.g. visual or tactile, in order to modify its quality or its intelligibility
    • G10L21/02—Speech enhancement, e.g. noise reduction or echo cancellation
    • G10L21/0208—Noise filtering
    • G10L21/0216—Noise filtering characterised by the method used for estimating noise
    • G10L2021/02161—Number of inputs available containing the signal or the noise to be suppressed
    • G10L2021/02165—Two microphones, one receiving mainly the noise signal and the other one mainly the speech signal
    • G—PHYSICS
    • G10—MUSICAL INSTRUMENTS; ACOUSTICS
    • G10L—SPEECH ANALYSIS TECHNIQUES OR SPEECH SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING TECHNIQUES; SPEECH OR AUDIO CODING OR DECODING
    • G10L21/00—Speech or voice signal processing techniques to produce another audible or non-audible signal, e.g. visual or tactile, in order to modify its quality or its intelligibility
    • G10L21/02—Speech enhancement, e.g. noise reduction or echo cancellation
    • G10L21/0208—Noise filtering
    • G10L21/0216—Noise filtering characterised by the method used for estimating noise

Definitions

  • the invention relates to the technical field of medical measurement and signal processing, in particular to a dual microphone adaptive filtering algorithm and application for collecting body sound signals.
  • Remote auscultation allows users to enjoy telemedicine services without leaving home, making it possible to visit doctors efficiently anytime, anywhere, and greatly reducing the cost of follow-up consultations for patients with chronic diseases.
  • remote auscultation has high requirements on the anti-noise ability of the auscultation system: the weak body sound signal is extremely susceptible to interference from environmental noise.
  • the doctor does not understand the situation of the patient’s environment and it is difficult to judge the abnormality heard. The sound is the murmur of the patient's body sound or environmental noise, which is prone to misdiagnosis. For this reason, the remote auscultation system must take effective measures to suppress the interference of environmental noise.
  • a common method is to use piezoelectric thin-film pickups to collect body sound signals.
  • the piezoelectric film pickup collects displacement signals, so it is not easily interfered by environmental noise.
  • the auscultation head needs to have a specific structural design, and the cost is too high to be easily extended to home users.
  • One of the preferred sensors for electronic stethoscopes is the electret microphone pickup, which has the advantages of simple structural design, low cost, and wide dynamic range.
  • the electret microphone is very sensitive. Even if it is enclosed in a metal cavity, it can collect ambient noise.
  • a matching filtering method must be designed before it can be used for remote auscultation.
  • the diversity of remote auscultation applications greatly increases the difficulty of filtering method design: the environmental noise is complex and diverse, the frequency distribution is wide, and cannot be modeled, especially the distribution of noise and body sound signals such as speech and music sounds in time and frequency band Both may overlap, and traditional filtering methods are not easy to filter.
  • one of the most common methods for environmental noise filtering is dual-microphone adaptive filtering.
  • the main microphone is used to collect noisy body sound signals
  • the secondary microphone collects ambient noise
  • the environmental noise measured by the secondary microphone is linearly processed to cancel the noisy body. Noise in the audio signal to achieve denoising.
  • the reasonable value of the adaptive step size is the key to ensure the effect of adaptive filtering, but its adjustment is often time-consuming and difficult, which is very difficult.
  • the normalized least mean square algorithm is commonly used to set the adaptive step size according to the amplitude of the environmental noise. Usually only a limited number of adjustments to the adjustment factor are needed to achieve rapid convergence of the filter weights, which greatly reduces the adaptive filtering The difficulty of algorithm tuning.
  • an object of the present invention is to provide a method for collecting body sound signals, which can achieve rapid convergence of filter weights, can avoid signal distortion, and quickly and reliably suppress environmental noise interference
  • the dual microphone adaptive filtering algorithm is particularly suitable for electronic auscultation.
  • Another object of the present invention is to provide an application of the above-described dual microphone adaptive filtering algorithm for collecting body sound signals.
  • a dual-microphone adaptive filtering algorithm for collecting body sound signals characterized in that at least one main and one secondary microphones are used to collect signals; the main microphone It is used to collect noisy body sound signals, and the auxiliary microphone is used to collect ambient noise; the same high-pass filtering process is performed on the signal collected by the main microphone and the signal collected by the auxiliary microphone, so that the main microphone signal and the auxiliary The microphone signal has a good linear correlation; the main microphone signal and the secondary microphone signal after high-pass filtering are processed by the normalized least mean square algorithm to calculate the adaptive filter weight and the error signal to filter out the main microphone signal.
  • Environmental noise the first low-pass filtering process is performed on the error signal to restore the body sound signal, thereby obtaining the body sound signal output by the adaptive filtering algorithm.
  • At least one main and two microphones are used to collect signals; the main microphone is used to collect noisy body sound signals, and the sub microphone is used to collect ambient noise; the signal collected by the main microphone and the sub microphone are collected
  • the signals are processed by the same high-pass filter, so that the main and secondary microphone signals after high-pass filtering have a good linear correlation; the normalized minimum mean square is used for the main and sub microphone signals after high-pass filtering
  • the algorithm calculates the weight of the adaptive filter and calculates the error signal to filter out the environmental noise in the main microphone signal; the first low-pass filtering process is performed on the error signal to restore the body sound signal, so as to obtain the volume output by the adaptive filter algorithm
  • the sound signal refers to the following steps:
  • Step S2 Obtain the main microphone signal d(k) and sub microphone signal x(k) at the current moment;
  • Step S3 determine the size of the serial number k at the current moment:
  • step S4 the main microphone signal d(k) and the sub microphone signal x(k) are subjected to the same high-pass filtering process to obtain the main microphone signal after the high-pass filtering process And high-pass filtered secondary microphone signal To reduce the amplitude difference between the body sound signal and the ambient noise in the main microphone signal, so that the main microphone signal and the sub microphone signal after high-pass filtering have a higher degree of linear correlation;
  • step S5 the filter output y(k) is calculated:
  • Step S6 calculate the error signal e(k):
  • Step S7 calculate the adaptive step size normalization coefficient ⁇ (k);
  • Step S8 update the filter weight W(k+1,i):
  • ⁇ is the adjustment factor
  • Step S9 the first low-pass filtering process is performed on the error signal e(k) to restore the body sound signal, and the signal after the first low-pass filtering process is obtained
  • Step S10 output the output signal o(k) of the adaptive filter algorithm at the kth time; determine the adaptive filter termination indicator variable: if the adaptive filter termination indicator variable is true, end the adaptive filter algorithm, otherwise skip to step S2 for calculation The output of the adaptive filtering algorithm at the next moment.
  • the value range of the filter order M is: M ⁇ [10,200].
  • the high-pass filtering process adopts one of the following two schemes:
  • pulse transfer function G HP (z) of the high-pass filtering processor pulse transfer function G HP (z) of the cut-off frequency is f HPc range: 500 ⁇ 1200Hz;
  • the pulse transfer function G 1LP (z) 1/G HP (z) of the low-pass filter used in the first low-pass filtering process in step S9.
  • the value range of the adjustment factor ⁇ is: ⁇ [0.1,1].
  • outputting the output signal o(k) of the adaptive filtering algorithm at time k refers to: adopting one of the following two methods:
  • the signal after the first low-pass filtering process The second low-pass filtering process is performed to further suppress environmental noise interference, and the signal after the second low-pass filtering process is output as the output signal o(k) of the adaptive filtering algorithm at the k-th time.
  • the second low-pass filtering process uses a low-pass filter with a pulse transfer function of G 2LP (z), and the value of the cut-off frequency f LPc of the pulse transfer function G 2LP (z)
  • the range is: 1200 ⁇ 1600Hz.
  • An application of the above-mentioned dual microphone adaptive filtering algorithm for collecting body sound signals is characterized in that it is applied to electronic auscultation devices and/or electronic wearable devices, and uses the body sound signals output by the adaptive filtering algorithm as electronic auscultation devices and /Or the output signal of the electronic wearable device.
  • the electronic auscultation device and/or electronic wearable device can assist medical personnel to auscultate the patient; the electronic auscultation device can also remotely transmit the body sound signal output by the adaptive filtering algorithm to the auscultation system, and the auscultation system provides the received body sound signal to The medical staff conducts remote auscultation.
  • the medical staff can listen to the patient's body sound without meeting with the patient, which solves the technical problem of clear monitoring of body sound in order to realize remote medical treatment.
  • the algorithm of the present invention adds high-pass filter processing and the first low-pass filter processing.
  • the first heart sound and the second heart sound appearing periodically will cause periodic and large-scale adjustment of the filter parameters; wherein, the ordinate of Fig. 8(b) is the filter at the kth time
  • the adjustment range of the weight parameter, the adjustment range passes the 2-norm of the difference between the filter weight vectors of two adjacent moments, that is to measure.
  • body sound signals such as heart sounds, breath sounds, and bowel sounds are low-frequency signals, and their effective frequency bands fall from 0 to 1600 Hz, and most of their energy is concentrated in the low frequency bands below 500 Hz.
  • the use of high-pass filter processing helps to reduce the amplitude difference between the body sound signal s(k) and the environmental noise n(k) in the main microphone signal, and increase the environmental noise n(k) to the filter weight W( While reducing the impact of k+1,i), reduce the impact of the body sound signal s(k) on the filter weight W(k+1,i) (compare Figure 8(b) and Figure 8(c)
  • adaptive filtering uses the linear correlation between the environmental noise x(k) measured by the secondary microphone and the environmental noise n(k) measured by the main microphone to filter the environmental noise n( k);
  • the higher the degree of linear correlation between the two, the more significant the adaptive filter has on the suppression of environmental noise n(k). Since the body sound signal s(k) is linearly independent of the ambient noise n(k), this also means that the sub microphone signal x(k) and the main microphone signal d(k) s(k)+n(k)
  • High-pass filtering helps improve this correlation, versus The linear correlation coefficient between can be increased several times or even more than ten times than the linear correlation coefficient between x(k) and d(k), so the effect of adaptive filtering can be greatly improved.
  • the purpose of the first low-pass filtering process is to restore the body sound signal s(k), so the pulse transfer function of the first low-pass filtering process should be the inverse of the pulse transfer function of the high-pass filtering process.
  • a second low-pass filtering process can be introduced to Further suppress the interference of environmental noise.
  • the present invention has the following advantages and beneficial effects:
  • the invention preprocesses the main microphone signal and the sub microphone signal through high-pass filtering to improve the environmental noise measured by the sub microphone x(k) and the environmental noise measured by the main microphone n(k) The degree of linear correlation between them, and further low-pass filtering the processing results of the normalized least mean square algorithm, so as to achieve the purpose of quickly and reliably suppressing environmental noise interference; especially suitable for the field of electronic auscultation;
  • the algorithm of the present invention has a small calculation amount, and the filter convergence speed is fast while avoiding signal distortion, and the computing power requirement of the hardware device is low. It is particularly suitable for small wearable auscultation devices and small electronic stethoscopes, and the algorithm of the present invention is also suitable for It is used in the electronic auscultation auxiliary diagnosis and treatment system for hospitals and families.
  • FIG. 1 is a flowchart of the algorithm of the present invention
  • 3(a) to 3(f) are comparison charts of the amplitude and frequency spectrum of the noisy body sound signal, the noisy body sound signal after the high-pass filtering process, and the output signal of the adaptive filtering algorithm in the present invention
  • 4(a) to 4(d) are comparison diagrams of amplitude and frequency spectrum of a noisy body sound signal before and after adaptive filtering in the present invention
  • 5(a) to 5(f) are comparison diagrams of the amplitude and frequency spectrum of the noisy body sound signal, the noisy body sound signal after the high-pass filtering process, and the environmental noise signal measured by the secondary microphone in the present invention
  • Fig. 8 (a) to (c) are the comparison curves of the filter parameter adjustment amplitude with time of the noisy body sound signal, the traditional adaptive filtering method in the present invention, and the filter parameter adjustment amplitude of the adaptive filtering method of the present invention with time.
  • a dual-microphone adaptive filtering algorithm for collecting body sound signals is shown in FIG. 1 and its principle is shown in FIG. 2.
  • At least one main and one pair of two microphones are used to collect signals; the main microphone is used to Collect noisy body sound signals, and the secondary microphone is used to collect ambient noise; perform the same high-pass filtering process on the signal collected by the main microphone and the signal collected by the secondary microphone, so that the main microphone signal and the secondary microphone signal after the high-pass filtering process Good linear correlation; normalized least mean square algorithm is used to calculate the adaptive filter weight and error signal for the main microphone signal and the secondary microphone signal after high-pass filtering to filter out the environmental noise in the main microphone signal ; Perform the first low-pass filtering process on the error signal to restore the body sound signal, so as to obtain the body sound signal output by the adaptive filtering algorithm.
  • M is the filter order; the filter order M
  • the value range is preferably: M ⁇ [10,200];
  • Step S2 Obtain the main microphone signal d(k) and sub microphone signal x(k) at the current moment;
  • Step S3 determine the size of the serial number k at the current moment:
  • step S4 the main microphone signal d(k) and the sub microphone signal x(k) are subjected to the same high-pass filtering process to obtain the main microphone signal after the high-pass filtering process And high-pass filtered secondary microphone signal To reduce the amplitude difference between the body sound signal and the ambient noise in the main microphone signal, so that the main microphone signal and the sub microphone signal after high-pass filtering have a higher degree of linear correlation;
  • the high-pass filtering process preferably adopts one of the following two schemes:
  • pulse transfer function G HP (z) of the high-pass filtering processor pulse transfer function G HP (z) of the cut-off frequency is f HPc range: 500 ⁇ 1200Hz;
  • step S5 the filter output y(k) is calculated:
  • Step S6 calculate the error signal e(k):
  • Step S7 calculate the adaptive step size normalization coefficient ⁇ (k);
  • Step S8 update the filter weight W(k+1,i):
  • ⁇ is the adjustment factor
  • the value range of the adjustment factor ⁇ is preferably: ⁇ [0.1,1]
  • Step S9 the first low-pass filtering process is performed on the error signal e(k) to restore the body sound signal, and the signal after the first low-pass filtering process is obtained
  • Step S10 the signal after the first low-pass filtering process Output as the output signal o(k) of the adaptive filtering algorithm at the kth time; determine the adaptive filtering termination indicator variable: if the adaptive filtering termination indicator variable is true, end the adaptive filtering algorithm, otherwise skip to step S2 for calculation The output of the adaptive filtering algorithm for a moment.
  • An application of the above-mentioned dual microphone adaptive filtering algorithm for collecting body sound signals is characterized in that it is applied to electronic auscultation devices and/or electronic wearable devices, and uses the body sound signals output by the adaptive filtering algorithm as electronic auscultation devices and /Or the output signal of the electronic wearable device.
  • the electronic auscultation device and/or electronic wearable device can assist medical personnel to auscultate the patient; the electronic auscultation device can also remotely transmit the body sound signal output by the adaptive filtering algorithm to the auscultation system, and the auscultation system provides the received body sound signal to The medical staff conducts remote auscultation.
  • the medical staff can listen to the patient's body sound without meeting with the patient, which solves the technical problem of clear monitoring of body sound in order to realize remote medical treatment.
  • the algorithm of the present invention adds high-pass filter processing and the first low-pass filter processing.
  • the first heart sound and the second heart sound appearing periodically will cause periodic and large-scale adjustment of the filter parameters; wherein, the ordinate of Fig. 8(b) is the filter at the kth time
  • the adjustment range of the weight parameter, the adjustment range passes the 2-norm of the difference between the filter weight vectors of two adjacent moments, that is to measure.
  • body sound signals such as heart sounds, breath sounds, and bowel sounds are low-frequency signals, and their effective frequency bands fall from 0 to 1600 Hz, and most of their energy is concentrated in the low frequency bands below 500 Hz.
  • the use of high-pass filter processing helps to reduce the amplitude difference between the body sound signal s(k) and the environmental noise n(k) in the main microphone signal, and increase the environmental noise n(k) to the filter weight W( While reducing the impact of k+1,i), reduce the impact of the body sound signal s(k) on the filter weight W(k+1,i) (compare Figure 8(b) and Figure 8(c)
  • adaptive filtering uses the linear correlation between the environmental noise x(k) measured by the secondary microphone and the environmental noise n(k) measured by the main microphone to filter the environmental noise n( k);
  • the higher the degree of linear correlation between the two, the more significant the adaptive filter has on the suppression of environmental noise n(k). Since the body sound signal s(k) is linearly independent of the ambient noise n(k), this also means that the sub microphone signal x(k) and the main microphone signal d(k) s(k)+n(k)
  • High-pass filter processing helps to improve this correlation.
  • the linear correlation coefficient between can be increased several times or even more than ten times than the linear correlation coefficient between x(k) and d(k), so the effect of adaptive filtering can be greatly improved.
  • the purpose of the first low-pass filtering process is to restore the body sound signal s(k), so the pulse transfer function of the first low-pass filtering process should be the inverse of the pulse transfer function of the high-pass filtering process.
  • Figures 3(a) to 3(f) are the comparison of the amplitude and frequency spectrum of the noisy body sound signal, the noisy body sound signal after high-pass filtering, and the output signal of the adaptive filtering algorithm; among them, Figure 3(a) is Figure 3(b) is the spectrum diagram of noisy body sound signal, Figure 3(c) is the diagram of noisy body sound signal after high-pass filtering, and Figure 3(d) is noisy after high-pass filtering Spectral diagram of body sound signal, Figure 3(e) is the output signal diagram of the adaptive filtering algorithm, and Figure 3(f) is the output signal spectrum diagram of the adaptive filtering algorithm.
  • Figures 4(a) to 4(d) are comparison charts of the amplitude and frequency spectrum of the noisy body sound signal before and after adaptive filtering; of which, Figure 4(a) is the noisy body sound signal diagram, and Figure 4(b) Figure 4(c) is the output signal graph of the adaptive filtering algorithm, and Figure 4(d) is the output signal spectral graph of the adaptive filtering algorithm. It can be seen from the figure that the environmental noise is greatly suppressed after adaptive filtering.
  • Figures 5(a) to 5(f) are comparison charts of the amplitude and frequency spectrum of the noisy body sound signal, the noisy body sound signal after the high-pass filtering process, and the environmental noise signal measured by the secondary microphone; of which, Figure 5(a ) Is a diagram of a noisy body sound signal, FIG. 5(b) is a spectrum diagram of a noisy body sound signal, FIG. 5(c) is a diagram of a noisy body sound signal after high-pass filtering, and FIG. 5(d) is after a high pass filtering process Spectral diagram of noisy body sound signal.
  • Figure 5(e) is the sub-microphone signal diagram after the high-pass filtering process.
  • Figure 5(f) is the sub-microphone signal spectrum diagram after the high-pass filtering process.
  • the high-pass filtering process has a noisy body
  • the amplitude of the low frequency band of the audio signal is greatly reduced.
  • the main microphone signal and the sub microphone signal are more correlated, which helps to improve the adaptive filtering effect.
  • Fig. 8 (a) to (c) are the comparison curves of the filter parameter adjustment amplitude with time of the noisy body sound signal, the traditional adaptive filtering method in the present invention, and the filter parameter adjustment amplitude of the adaptive filtering method of the present invention with time.
  • the dual-microphone adaptive filtering algorithm used to collect body sound signals includes the following steps:
  • Step S2 Obtain the main microphone signal d(k) and sub microphone signal x(k) at the current moment;
  • the linear correlation coefficient between can be increased several times or even more than ten times than the linear correlation coefficient between x(k) and d(k), which can greatly improve the effect of adaptive filtering;
  • step S5 the filter output y(k) is calculated:
  • Step S6 calculate the error signal e(k):
  • Step S7 Calculate the adaptive step size normalization coefficient ⁇ (k):
  • Step S8 update the filter weight W(k+1,i):
  • step S9 the pre-emphasis processing (ie, the first low-pass filter processing) is performed on the error signal e(k) to obtain
  • Step S10 the signal after the first low-pass filtering process Output as the output signal o(k) of the adaptive filtering algorithm at the kth time; determine the adaptive filtering termination indicator variable: if the adaptive filtering termination indicator variable is true, end the adaptive filtering algorithm, otherwise skip to step S2 for calculation The output of the adaptive filtering algorithm for a moment.
  • the adaptive filter termination indicator variable is obtained by reading the stop button. When the stop button information is pressed, the adaptive filter termination indicator variable is set to true, otherwise it is set to false.
  • step S10 Signal after the first low-pass filtering
  • the second low-pass filtering process is performed to further suppress environmental noise interference, and the signal after the second low-pass filtering process is output as the output signal o(k) of the adaptive filtering algorithm at the k-th time.
  • body sound signals are low-frequency signals
  • a second low-pass filtering process can be introduced to Further suppress the interference of environmental noise.
  • the remaining steps in this embodiment are the same as those in Embodiment 1.
  • the second low-pass filter processing using a pulse transfer function G 2LP (z) of the low-pass filter ranges: 1200 ⁇ 1600Hz.
  • step S10 the signal after the first low-pass filtering process For the second low-pass filtering process, the result is the output signal o(k) of the adaptive filtering algorithm at time k:
  • the order m LP can choose 4-8 or higher order, parameters with Determined by the cut-off frequency f LPc (taken as 1500Hz ) and the sampling frequency f s (can be determined using the Butterworth low-pass filter design algorithm);
  • the adaptive filtering termination indicator variable is judged: if the adaptive filtering termination indicator variable is true, the adaptive filtering algorithm is ended, otherwise skip to step S2 to calculate the output of the adaptive filtering algorithm at the next moment.
  • steps S4 and S9 in this embodiment are different from the specific example in Embodiment 1: In this embodiment,
  • step S4 the main microphone signal d(k) and the sub microphone signal x(k) are subjected to the same high-pass filtering process to obtain the main microphone signal after the high-pass filtering process
  • high-pass filtered secondary microphone signal High-pass filtering a pulse transfer function G HP (z) of the high-pass filtering processor, pulse transfer function G HP (z) of the cut-off frequency f HPc ranges: 500 ⁇ 1200Hz.
  • the following formula is used for high-pass filtering:
  • the order m HP can choose 2 ⁇ 8 or higher order, parameter with It is determined by the cut-off frequency f HPc (taken as 500 Hz) and the sampling frequency f s (which can be determined by using the Butterworth high-pass filter design algorithm). If the passband band gain is higher than 20 dB, the effect is better.
  • Step S9 the first low-pass filtering process is performed on the error signal e(k) to obtain the signal after the first low-pass filtering process

Landscapes

  • Engineering & Computer Science (AREA)
  • Acoustics & Sound (AREA)
  • Health & Medical Sciences (AREA)
  • Physics & Mathematics (AREA)
  • Signal Processing (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • General Health & Medical Sciences (AREA)
  • Medical Informatics (AREA)
  • Public Health (AREA)
  • Heart & Thoracic Surgery (AREA)
  • Otolaryngology (AREA)
  • Molecular Biology (AREA)
  • Surgery (AREA)
  • Animal Behavior & Ethology (AREA)
  • Biomedical Technology (AREA)
  • Veterinary Medicine (AREA)
  • Computational Linguistics (AREA)
  • Quality & Reliability (AREA)
  • Audiology, Speech & Language Pathology (AREA)
  • Human Computer Interaction (AREA)
  • Multimedia (AREA)
  • Circuit For Audible Band Transducer (AREA)
  • Soundproofing, Sound Blocking, And Sound Damping (AREA)

Abstract

一种用于采集体音信号的双麦克风自适应滤波方法，该方法包括：采用至少一主一副两个麦克风来采集信号，主麦克风用以采集带噪体音信号，副麦克风用以采集环境噪音；对主麦克风采集到的信号和副麦克风采集到的信号作相同的高通滤波处理；对高通滤波处理后的主麦克风信号和副麦克风信号采用归一化最小均方算法计算自适应滤波器权值并计算误差信号，以滤除主麦克风信号中的环境噪音；对误差信号作第一次低通滤波处理以复原体音信号，从而得到自适应滤波算法输出的体音信号。该方法既可实现滤波器权值快速收敛，又可避免信号失真，快速可靠地抑制环境噪声干扰。

Description

用于采集体音信号的双麦克风自适应滤波算法及应用 技术领域
本发明涉及医学测量和信号处理技术领域，尤其涉及一种用于采集体音信号的双麦克风自适应滤波算法及应用。
背景技术
远程听诊令使用者足不出户即可享受远程医疗服务，使随时随地高效率就诊成为可能，极大地减轻慢性病患者随诊的成本。但远程听诊对听诊系统的抗噪能力要求很高：微弱的体音信号极易受到环境噪声的干扰，而远程听诊过程中，医生不了解患者所处环境的情况，难以判断所听到的异常声音为患者体音的杂音还是环境噪声，容易出现误诊。为此，远程听诊系统必须采取有效措施抑制环境噪声的干扰。
一种常见的方法是采用压电薄膜拾音器采集体音信号。压电薄膜拾音器采集的是位移信号，故不易受到环境噪声的干扰。但为保证灵敏度，其听诊头需进行特定的结构设计，成本偏高而不易推广至家庭用户。
电子听诊器的首选传感器之一是驻极体麦克风拾音器，其具有结构设计简单、成本低、动态范围宽等优点。但驻极体麦克风非常灵敏，即使封装于金属腔体内亦能采集到环境噪音，须设计配套的滤波方法方可用于远程听诊。但是远程听诊应用场合的多样性极大地增加了滤波方法设计的难度：环境噪声复杂多样，频率分布广，且无法建模，尤其是语音和音乐声等噪音与体音信号在时间分布和频段分布上均可能重叠，传统滤波方法不易滤除。
目前对环境噪声滤波最常见的方法之一为双麦克风自适应滤波，用主麦克风采集带噪体音信号，副麦克风采集环境噪音，副麦克风所测环境噪音作线性处理后用于抵消带噪体音信号中的噪音而实现去噪。自适应步长的合理取值是保证自适应滤波效果的关键，但其调节往往费时费力，难度很大。目前常用归一化最小均方算法依据环境噪声的幅值设定自适应步长，通常仅需对调节因子作有限几次调整即可实现滤波器权值快速收敛，极大地降低了自适应滤波算法调参的难度。
在归一化最小均方算法的应用中，调节因子η的合理取值非常重要：滤波器权值迭代依赖于调节因子η：W(k+1,i)＝W(k,i)+η(d(k)-y(k))/ε(k)；其中，d(k)＝s(k)+n(k)，s(k)和n(k)分别为第k时刻的体音信号和环境噪声；而自 适应滤波器输出为：e(k)＝d(k)-y(k)；若调节因子过小则收敛慢，长时间无法达到抑制环境噪声的目的；而若调节因子过大则容易导致滤波器权值W(k+1,i)失调。
若将传统归一化最小均方算法应用于听诊滤波中，当体音信号s(k)的幅值远大于环境噪音n(k)的幅值时，第一、第二心音幅值很大，极易因调节因子η取值过大导致自适应滤波器参数失调而致使输出失真。若为降低信号失真程度而选取很小的调节因子η值，又会使得滤波器权值收敛过慢，失去实际应用的价值。信号保真和快速收敛之间的矛盾在通用的归一化最小均方算法中难以克服。
因此，双麦克风自适应滤波算法目前仍然不能直接应用在电子听诊上。
发明内容
为克服现有技术中的缺点与不足，本发明的一个目的在于提供一种用于采集体音信号、既可实现滤波器权值快速收敛、又可避免信号失真、快速可靠地抑制环境噪声干扰的双麦克风自适应滤波算法；该算法尤其适用于电子听诊。本发明的另一个目的在于提供一种上述用于采集体音信号的双麦克风自适应滤波算法的应用。
为了达到上述目的，本发明通过下述技术方案予以实现：一种用于采集体音信号的双麦克风自适应滤波算法，其特征在于：采用至少一主一副两个麦克风来采集信号；主麦克风用以采集带噪体音信号，副麦克风用以采集环境噪音；对主麦克风采集到的信号和副麦克风采集到的信号作相同的高通滤波处理，以使高通滤波处理后的主麦克风信号和副麦克风信号具有良好的线性相关程度；对高通滤波处理后的主麦克风信号和副麦克风信号采用归一化最小均方算法计算自适应滤波器权值并计算误差信号，以滤除主麦克风信号中的环境噪音；对误差信号作第一次低通滤波处理以复原体音信号，从而得到自适应滤波算法输出的体音信号。
优选地，所述的采用至少一主一副两个麦克风来采集信号；主麦克风用以采集带噪体音信号，副麦克风用以采集环境噪音；对主麦克风采集到的信号和副麦克风采集到的信号作相同的高通滤波处理，以使高通滤波处理后的主麦克风信号和副麦克风信号具有良好的线性相关程度；对高通滤波处理后的主麦克风信号和副麦克风信号采用归一化最小均方算法计算自适应滤波器权值并计算误差信号，以滤除主麦克风信号中的环境噪音；对误差信号作第一次低通滤波处理以复原体音信号，从而得到自适应滤波算法输出的体音信号，是指：包括如下步骤：
S1步，初始化当前时刻序号k＝0，滤波器权值W(0,i)＝0，i＝0，...，M-1，其中M为滤波器阶数；
S2步，获取当前时刻的主麦克风信号d(k)和副麦克风信号x(k)；
S3步，判断当前时刻序号k的大小：
若k＜M，则取第一次低通滤波处理后信号为
Figure PCTCN2019110287-appb-000001
同时令W(k,i)＝W(k-1,i)，并转至S10步；
若k≥M，则转至S4步；
S4步，对主麦克风信号d(k)和副麦克风信号x(k)作相同的高通滤波处理，得到高通滤波处理后的主麦克风信号
Figure PCTCN2019110287-appb-000002
和高通滤波处理后的副麦克风信号
Figure PCTCN2019110287-appb-000003
以缩小主麦克风信号中体音信号与环境噪音之间的幅值差距，从而使高通滤波处理后的主麦克风信号和副麦克风信号具有更高的线性相关程度；
S5步，计算滤波器输出y(k)：
Figure PCTCN2019110287-appb-000004
S6步，计算误差信号e(k)：
Figure PCTCN2019110287-appb-000005
S7步，计算自适应步长归一化系数ε(k)；
Figure PCTCN2019110287-appb-000006
其中，ζ为防止ε(k)＝0的正数；
S8步，更新滤波器权值W(k+1,i)：
W(k+1,i)＝W(k,i)+ηe(k)/ε(k)；
其中，η为调节因子；
S9步，对误差信号e(k)作第一次低通滤波处理以复原体音信号，得到第一次低通滤波处理后信号
Figure PCTCN2019110287-appb-000007
S10步，输出第k时刻自适应滤波算法的输出信号o(k)；判断自适应滤波终止指示变量：若自适应滤波终止指示变量为真，则结束自适应滤波算法，否则跳至S2步计算下一时刻自适应滤波算法的输出。
优选地，所述S1步中，滤波器阶数M的取值范围为：M∈[10,200]。
优选地，所述S4步中，高通滤波处理采用如下两种方案之一：
一、采用脉冲传递函数为G HP(z)的高通滤波处理器，脉冲传递函数G HP(z)的截止频率f HPc的取值范围为：500～1200Hz；
二、采用由m HP个一阶预加重环节1-α jz -1，j＝1，...，m HP，α j∈[0.9,1)串联而构成的预加重高通滤波器。
相应地，所述第一种方案中，S9步第一次低通滤波处理采用的低通滤波器的脉冲传递函数G 1LP(z)＝1/G HP(z)。
优选地，所述S8步中，调节因子η的取值范围为：η∈[0.1,1]。
优选地，所述S10步中，输出第k时刻自适应滤波算法的输出信号o(k)，是指：采用如下两种方式之一：
一、将第一次低通滤波处理后信号
Figure PCTCN2019110287-appb-000008
作为第k时刻自适应滤波算法的输出信号o(k)来输出；
二、对第一次低通滤波处理后信号
Figure PCTCN2019110287-appb-000009
作第二次低通滤波处理以进一步抑制环境噪音干扰，以第二次低通滤波处理后信号作为第k时刻自适应滤波算法的输出信号o(k)来输出。
优选地，所述第二种方式中，第二次低通滤波处理采用脉冲传递函数为G 2LP(z)的低通滤波器，脉冲传递函数G 2LP(z)的截止频率f LPc的取值范围为：1200～1600Hz。
一种上述用于采集体音信号的双麦克风自适应滤波算法的应用，其特征在于：应用于电子听诊设备和/或电子穿戴设备，将自适应滤波算法输出的体音信号作为电子听诊设备和/或电子穿戴设备的输出信号。电子听诊设备和/或电子穿戴设备可辅助医疗人员对病人进行听诊；电子听诊设备还可以将自适应滤波算法输出的体音信号远程传输至听诊系统，听诊系统将接收到的体音信号提供给医疗人员进行远程听诊，医疗人员不需要与病人会面也可听取病人体音，为实现远程看病解决了体音清晰监听这一技术难题。
本发明算法的技术原理是：
与传统归一化最小均方算法相比较，本发明算法增加了高通滤波处理和第一次低通滤波处理。
在心音听诊中，第一心音和第二心音的幅值往往远高于环境噪声。因而导致滤波偏差e(k)＝d(k)-y(k)在滤波器参数收敛的过程中周期性增大，并进而引起滤波器参数周期性失调。如图8(b)所示，周期性出现的第一心音和第二心音会导致滤波器参数出现周期性的大幅度调整；其中，图8(b)的纵坐标为第k时刻滤波器权值参数的调整幅度，该调整幅度通过相邻两时刻滤波器权值向量之差的2范数，即
Figure PCTCN2019110287-appb-000010
来衡量。
相对于常见的语音等环境噪音，心音、呼吸音和肠鸣音等体音信号属低频信号，其有效频段落于0～1600Hz，且其绝大部分能量集中于500Hz以下的低频段。采用高通滤波处理，有助于缩小主麦克风信号中体音信号s(k)与环境噪音n(k)之间的幅值差距，在加大环境噪音n(k)对滤波器权值W(k+1,i)的影响的同时，减小体音信号s(k)对滤波器权值W(k+1,i)的影响(可对比图8(b)和图8(c)中||ΔW(k)|| 2的幅值)，从而达到降低自适应滤波器输出失真程度的目的，同时也降低了调节因子η调节的难度。
其原理亦可解释为：自适应滤波利用副麦克风所测环境噪音x(k)与主麦克风所测环境噪音n(k)之间的线性相关性来滤除主麦克风信号中的环境噪音n(k)；两者的线性相关程度越高，自适应滤波对环境噪音n(k)的抑制效果越显著。由于体音信号s(k)与环境噪音n(k)线性无关，这也意味着副麦克风信号x(k)与主 麦克风信号d(k)＝s(k)+n(k)之间的线性相关程度越高，自适应滤波效果越佳。高通滤波处理有助于提高这种相关性，
Figure PCTCN2019110287-appb-000011
与
Figure PCTCN2019110287-appb-000012
之间的线性相关系数可比x(k)与d(k)之间的线性相关系数增大数倍甚至十倍以上，因而可极大地改善自适应滤波的效果。
第一次低通滤波处理目的是复原体音信号s(k)，故第一次低通滤波处理的脉冲传递函数应为高通滤波处理的脉冲传递函数的倒数。
考虑到相对于大部分环境噪音而言，体音信号属低频信号，在第一次低通滤波处理后，在得到自适应滤波算法的输出信号前，可引入第二次低通滤波处理，以进一步抑制环境噪音的干扰。
与现有技术相比，本发明具有如下优点与有益效果：
1、本发明针对体音信号频段范围特点，通过高通滤波对主麦克风信号和副麦克风信号进行预处理，来提高副麦克风所测环境噪音x(k)与主麦克风所测环境噪音n(k)之间的线性相关程度，并对归一化最小均方算法的处理结果进一步做低通滤波，从而达到快速、可靠地抑制环境噪声干扰的目的；尤其适用于电子听诊技术领域；
2、本发明算法计算量小，且在避免信号失真的同时滤波器收敛速度快，对硬件设备的计算能力要求低，特别适用于小型可穿戴听诊设备和小型电子听诊器，同时本发明算法也适合在医院和家庭用电子听诊辅助诊疗系统中应用。
附图说明
图1为本发明算法的流程图；
图2为本发明算法的原理图；
图3(a)至图3(f)为本发明中带噪体音信号、高通滤波处理后带噪体音信号、自适应滤波算法输出信号的幅值和频谱对比图；
图4(a)至图4(d)为本发明中带噪体音信号在自适应滤波前后的幅值和频谱对比图；
图5(a)至图5(f)为本发明中带噪体音信号、高通滤波处理后带噪体音信号、副麦克风所测环境噪音信号的幅值和频谱对比图；
图6为实施例二算法的流程图；
图7为实施例二算法的原理图；
图8(a)至(c)为本发明中带噪体音信号、传统自适应滤波方法滤波器参数调整幅度随时间变化曲线、本发明自适应滤波方法滤波器参数调整幅度随时间变化曲线对比图。
具体实施方式
下面结合附图与具体实施方式对本发明作进一步详细的描述。
实施例一
本实施例用于采集体音信号的双麦克风自适应滤波算法，其流程如图1所示，其原理如图2所示，采用至少一主一副两个麦克风来采集信号；主麦克风用以采集带噪体音信号，副麦克风用以采集环境噪音；对主麦克风采集到的信号和副麦克风采集到的信号作相同的高通滤波处理，以使高通滤波处理后的主麦克风信号和副麦克风信号具有良好的线性相关程度；对高通滤波处理后的主麦克风信号和副麦克风信号采用归一化最小均方算法计算自适应滤波器权值并计算误差信号，以滤除主麦克风信号中的环境噪音；对误差信号作第一次低通滤波处理以复原体音信号，从而得到自适应滤波算法输出的体音信号。
具体地说，包括如下步骤：
S1步，初始化当前时刻序号k＝0，滤波器权值W(0,i)＝0，i＝0，...，M-1，其中M为滤波器阶数；滤波器阶数M的取值范围优选为：M∈[10,200]；
S2步，获取当前时刻的主麦克风信号d(k)和副麦克风信号x(k)；
S3步，判断当前时刻序号k的大小：
若k＜M，则取第一次低通滤波处理后信号为
Figure PCTCN2019110287-appb-000013
同时令W(k,i)＝W(k-1,i)，并转至S10步；
若k≥M，则转至S4步；
S4步，对主麦克风信号d(k)和副麦克风信号x(k)作相同的高通滤波处理，得到高通滤波处理后的主麦克风信号
Figure PCTCN2019110287-appb-000014
和高通滤波处理后的副麦克风信号
Figure PCTCN2019110287-appb-000015
以缩小主麦克风信号中体音信号与环境噪音之间的幅值差距，从而使高通滤波处理后的主麦克风信号和副麦克风信号具有更高的线性相关程度；
高通滤波处理优选采用如下两种方案之一：
一、采用脉冲传递函数为G HP(z)的高通滤波处理器，脉冲传递函数G HP(z)的截止频率f HPc的取值范围为：500～1200Hz；
二、采用由m HP个一阶预加重环节1-α jz -1，j＝1，...，m HP，α j∈[0.9,1)串联而构成的预加重高通滤波器；
S5步，计算滤波器输出y(k)：
Figure PCTCN2019110287-appb-000016
S6步，计算误差信号e(k)：
Figure PCTCN2019110287-appb-000017
S7步，计算自适应步长归一化系数ε(k)；
Figure PCTCN2019110287-appb-000018
其中，ζ为防止ε(k)＝0、很小的正数，例如取ζ＝10 -5；
S8步，更新滤波器权值W(k+1,i)：
W(k+1,i)＝W(k,i)+ηe(k)/ε(k)；
其中，η为调节因子；调节因子η的取值范围优选为：η∈[0.1,1]；
S9步，对误差信号e(k)作第一次低通滤波处理以复原体音信号，得到第一次低通滤波处理后信号
Figure PCTCN2019110287-appb-000019
在S4步的高通滤波处理采用第一种方案时，S9步第一次低通滤波处理采用的低通滤波器的脉冲传递函数G 1LP(z)＝1/G HP(z)。
S10步，将第一次低通滤波处理后信号
Figure PCTCN2019110287-appb-000020
作为第k时刻自适应滤波算法的输出信号o(k)来输出；判断自适应滤波终止指示变量：若自适应滤波终止指示变量为真，则结束自适应滤波算法，否则跳至S2步计算下一时刻自适应滤波算法的输出。
一种上述用于采集体音信号的双麦克风自适应滤波算法的应用，其特征在于：应用于电子听诊设备和/或电子穿戴设备，将自适应滤波算法输出的体音信号作为电子听诊设备和/或电子穿戴设备的输出信号。电子听诊设备和/或电子穿戴设备可辅助医疗人员对病人进行听诊；电子听诊设备还可以将自适应滤波算法输出的体音信号远程传输至听诊系统，听诊系统将接收到的体音信号提供给医疗人员进行远程听诊，医疗人员不需要与病人会面也可听取病人体音，为实现远程看病解决了体音清晰监听这一技术难题。
本发明算法的技术原理是：
与通用的归一化最小均方算法相比较，本发明算法增加了高通滤波处理和第一次低通滤波处理。
在心音听诊中，第一心音和第二心音的幅值往往远高于环境噪声。因而导致滤波偏差e(k)＝d(k)-y(k)在滤波器参数收敛的过程中周期性增大，并进而引起滤波器参数周期性失调。如图8(b)所示，周期性出现的第一心音和第二心音会导致滤波器参数出现周期性的大幅度调整；其中，图8(b)的纵坐标为第k时刻滤波器权值参数的调整幅度，该调整幅度通过相邻两时刻滤波器权值向量之差的2范数，即
Figure PCTCN2019110287-appb-000021
来衡量。
相对于常见的语音等环境噪音，心音、呼吸音和肠鸣音等体音信号属低频信号，其有效频段落于0～1600Hz，且其绝大部分能量集中于500Hz以下的低频段。采用高通滤波处理，有助于缩小主麦克风信号中体音信号s(k)与环境噪音n(k)之间的幅值差距，在加大环境噪音n(k)对滤波器权值W(k+1,i)的影响的同时，减小体音信号s(k)对滤波器权值W(k+1,i)的影响(可对比图8(b)和图8(c)中||ΔW(k)|| 2的幅值)，从而达到降低自适应滤波器输出失真程度的目的，同时也降低了调节因子η调节的难度。
其原理亦可解释为：自适应滤波利用副麦克风所测环境噪音x(k)与主麦克风所测环境噪音n(k)之间的线性相关性来滤除主麦克风信号中的环境噪音n(k)； 两者的线性相关程度越高，自适应滤波对环境噪音n(k)的抑制效果越显著。由于体音信号s(k)与环境噪音n(k)线性无关，这也意味着副麦克风信号x(k)与主麦克风信号d(k)＝s(k)+n(k)之间的线性相关程度越高，自适应滤波效果越佳。高通滤波处理有助于提高这种相关性，当采用二阶以上所述预加重处理后，
Figure PCTCN2019110287-appb-000022
与
Figure PCTCN2019110287-appb-000023
之间的线性相关系数可比x(k)与d(k)之间的线性相关系数增大数倍甚至十倍以上，因而可极大地改善自适应滤波的效果。
第一次低通滤波处理目的是复原体音信号s(k)，故第一次低通滤波处理的脉冲传递函数应为高通滤波处理的脉冲传递函数的倒数。
图3(a)至图3(f)为带噪体音信号、高通滤波处理后带噪体音信号、自适应滤波算法输出信号的幅值和频谱对比图；其中，图3(a)为带噪体音信号图，图3(b)为带噪体音信号频谱图，图3(c)为高通滤波处理后带噪体音信号图，图3(d)为高通滤波处理后带噪体音信号频谱图，图3(e)为自适应滤波算法输出信号图，图3(f)为自适应滤波算法输出信号频谱图。
图4(a)至图4(d)为带噪体音信号在自适应滤波前后的幅值和频谱对比图；其中，图4(a)为带噪体音信号图，图4(b)为带噪体音信号频谱图，图4(c)为自适应滤波算法输出信号图，图4(d)为自适应滤波算法输出信号频谱图。由图可见，自适应滤波后环境噪音被大大抑制。
图5(a)至图5(f)为带噪体音信号、高通滤波处理处理后带噪体音信号、副麦克风所测环境噪音信号的幅值和频谱对比图；其中，图5(a)为带噪体音信号图，图5(b)为带噪体音信号频谱图，图5(c)为高通滤波处理后带噪体音信号图，图5(d)为高通滤波处理后带噪体音信号频谱图，图5(e)为高通滤波处理后副麦克风信号图，图5(f)为高通滤波处理后副麦克风信号频谱图；由图可见，高通滤波处理后带噪体音信号低频段幅值大大减少，高通滤波处理后主麦克风信号和副麦克风信号相关程度更高，有助于改善自适应滤波效果。
图8(a)至(c)为本发明中带噪体音信号、传统自适应滤波方法滤波器参数调整幅度随时间变化曲线、本发明自适应滤波方法滤波器参数调整幅度随时间变化曲线对比图；其中，图8(b)所示的、传统自适应滤波方法所得||ΔW(k)|| 2的幅值会因第一第二心音的周期性出现而周期性突变，导致滤波器参数周期性失调；采用本发明自适应滤波方法后，所得||ΔW(k)|| 2的幅值不再因第一第二心音的周期性出现而周期性突变，克服了周期性失调现象，改善了滤波器参数收敛性能。
下面以具体示例进行说明：
用于采集体音信号的双麦克风自适应滤波算法包括以下步骤：
S1步，初始化：令当前时刻序号k＝0，滤波器权值W(0,i)＝0，i＝0，...，19，即 取滤波器阶数为20；
S2步，获取当前时刻的主麦克风信号d(k)和副麦克风信号x(k)；
S3步，判断当前时刻序号k的大小：若k＜20，则取第一次低通滤波处理后信号为
Figure PCTCN2019110287-appb-000024
同时令W(k,i)＝W(k-1,i)，i＝0,…,19，并转至S10步；
若k≥20，则转至S4步；
S4步，分别对主麦克风信号d(k)和副麦克风信号x(k)作相同的二阶预加重处理，其效果为高通滤波处理；即高通滤波处理采用由2个一阶预加重环节1-α jz -1，j＝1，2 j＝1,…,m HP，α j∈[0.9,1)串联而构成的预加重高通滤波器；
Figure PCTCN2019110287-appb-000025
Figure PCTCN2019110287-appb-000026
其中，α 1,α 2∈[0.95,1)；
所述预加重处理后，
Figure PCTCN2019110287-appb-000027
与
Figure PCTCN2019110287-appb-000028
之间的线性相关系数可比x(k)与d(k)之间的线性相关系数增大数倍甚至十倍以上，可极大地改善自适应滤波的效果；
S5步，计算滤波器输出y(k)：
Figure PCTCN2019110287-appb-000029
S6步，计算误差信号e(k)：
Figure PCTCN2019110287-appb-000030
S7步，计算自适应步长归一化系数ε(k)：
Figure PCTCN2019110287-appb-000031
S8步，更新滤波器权值W(k+1,i)：
W(k+1,i)＝W(k,i)+ηe(k)/ε(k)；
S9步，对误差信号e(k)作去预加重处理(即第一次低通滤波处理)得
Figure PCTCN2019110287-appb-000032
Figure PCTCN2019110287-appb-000033
S10步，将第一次低通滤波处理后信号
Figure PCTCN2019110287-appb-000034
作为第k时刻自适应滤波算法的 输出信号o(k)来输出；判断自适应滤波终止指示变量：若自适应滤波终止指示变量为真，则结束自适应滤波算法，否则跳至S2步计算下一时刻自适应滤波算法的输出。自适应滤波终止指示变量，是通过读取停止按键来获取，在按下停止按键信息时自适应滤波终止指示变量设定为真，否则设定为假。
实施例二
本实施例用于采集体音信号的双麦克风自适应滤波算法，其流程如图6所示，其原理如图7所示，与实施例一的区别在于：本实施例中，S10步，对第一次低通滤波处理后信号
Figure PCTCN2019110287-appb-000035
作第二次低通滤波处理以进一步抑制环境噪音干扰，以第二次低通滤波处理后信号作为第k时刻自适应滤波算法的输出信号o(k)来输出。考虑到相对于大部分环境噪音而言，体音信号属低频信号，在第一次低通滤波处理后，在得到自适应滤波算法的输出信号前，可引入第二次低通滤波处理，以进一步抑制环境噪音的干扰。本实施例的其余步骤与实施例一相同。
第二次低通滤波处理采用脉冲传递函数为G 2LP(z)的低通滤波器，脉冲传递函数G 2LP(z)的截止频率f LPc的取值范围为：1200～1600Hz。
相应地，在具体示例中，S10步，对第一次低通滤波处理后信号
Figure PCTCN2019110287-appb-000036
作第二次低通滤波处理，所得结果即为第k时刻自适应滤波算法的输出信号o(k)：
Figure PCTCN2019110287-appb-000037
其中，阶数m LP可选择4～8或更高阶数，参数
Figure PCTCN2019110287-appb-000038
和
Figure PCTCN2019110287-appb-000039
由截止频率f LPc(取为1500Hz)和采样频率f s共同决定(可采用巴特沃斯低通滤波器设计算法来确定)；
之后，判断自适应滤波终止指示变量：若自适应滤波终止指示变量为真，则结束自适应滤波算法，否则跳至S2步计算下一时刻自适应滤波算法的输出。
本实施例的其余步骤与实施例一相同。
实施例三
本实施例用于采集体音信号的双麦克风自适应滤波算法与实施例一的区别在于：本实施例中S4步和S9步与实施例一的具体示例不相同：本实施例中，
S4步，主麦克风信号d(k)和副麦克风信号x(k)作相同的高通滤波处理，得到高通滤波处理后的主麦克风信号
Figure PCTCN2019110287-appb-000040
和高通滤波处理后的副麦克风信号
Figure PCTCN2019110287-appb-000041
高通滤波处理采用脉冲传递函数为G HP(z)的高通滤波处理器，脉冲传递函数G HP(z)的截止频率f HPc的取值范围为：500～1200Hz。例如采用如下公式进行高通滤波处理：
Figure PCTCN2019110287-appb-000042
Figure PCTCN2019110287-appb-000043
其中，阶数m HP可选择2～8或更高阶数，参数
Figure PCTCN2019110287-appb-000044
和
Figure PCTCN2019110287-appb-000045
由截止频率f HPc(取为500Hz)和采样频率f s共同决定(可采用巴特沃斯高通滤波器设计算法来确定)，若能保证通带频段增益高于20dB则效果更佳。
S9步，对误差信号e(k)作第一次低通滤波处理，得到第一次低通滤波处理后信号
Figure PCTCN2019110287-appb-000046
Figure PCTCN2019110287-appb-000047
本实施例的其余步骤与实施例一相同。
上述实施例为本发明较佳的实施方式，但本发明的实施方式并不受上述实 施例的限制，其他的任何未背离本发明的精神实质与原理下所作的改变、修饰、替代、组合、简化，均应为等效的置换方式，都包含在本发明的保护范围之内。

Claims (9)

  1. 一种用于采集体音信号的双麦克风自适应滤波算法，其特征在于：采用至少一主一副两个麦克风来采集信号；主麦克风用以采集带噪体音信号，副麦克风用以采集环境噪音；对主麦克风采集到的信号和副麦克风采集到的信号作相同的高通滤波处理，以使高通滤波处理后的主麦克风信号和副麦克风信号具有良好的线性相关程度；对高通滤波处理后的主麦克风信号和副麦克风信号采用归一化最小均方算法计算自适应滤波器权值并计算误差信号，以滤除主麦克风信号中的环境噪音；对误差信号作第一次低通滤波处理以复原体音信号，从而得到自适应滤波算法输出的体音信号。
  2. 根据权利要求1所述的用于采集体音信号的双麦克风自适应滤波算法，其特征在于：所述的采用至少一主一副两个麦克风来采集信号；主麦克风用以采集带噪体音信号，副麦克风用以采集环境噪音；对主麦克风采集到的信号和副麦克风采集到的信号作相同的高通滤波处理，以使高通滤波处理后的主麦克风信号和副麦克风信号具有良好的线性相关程度；对高通滤波处理后的主麦克风信号和副麦克风信号采用归一化最小均方算法计算自适应滤波器权值并计算误差信号，以滤除主麦克风信号中的环境噪音；对误差信号作第一次低通滤波处理以复原体音信号，从而得到自适应滤波算法输出的体音信号，是指：包括如下步骤：
    S1步，初始化当前时刻序号k＝0，滤波器权值W(0,i)＝0，i＝0，...，M-1，其中M为滤波器阶数；
    S2步，获取当前时刻的主麦克风信号d(k)和副麦克风信号x(k)；
    S3步，判断当前时刻序号k的大小：
    若k＜M，则取第一次低通滤波处理后信号为
    Figure PCTCN2019110287-appb-100001
    同时令W(k,i)＝W(k-1,i)，并转至S10步；
    若k≥M，则转至S4步；
    S4步，对主麦克风信号d(k)和副麦克风信号x(k)作相同的高通滤波处理，得到高通滤波处理后的主麦克风信号
    Figure PCTCN2019110287-appb-100002
    和高通滤波处理后的副麦克风信号
    Figure PCTCN2019110287-appb-100003
    以缩小主麦克风信号中体音信号与环境噪音之间的幅值差距，从而使高通滤波处理后的主麦克风信号和副麦克风信号具有更高的线性相关程度；
    S5步，计算滤波器输出y(k)：
    Figure PCTCN2019110287-appb-100004
    S6步，计算误差信号e(k)：
    Figure PCTCN2019110287-appb-100005
    S7步，计算自适应步长归一化系数ε(k)；
    Figure PCTCN2019110287-appb-100006
    其中，ζ为防止ε(k)＝0的正数；
    S8步，更新滤波器权值W(k+1,i)：
    W(k+1,i)＝W(k,i)+ηe(k)/ε(k)；
    其中，η为调节因子；
    S9步，对误差信号e(k)作第一次低通滤波处理以复原体音信号，得到第一次低通滤波处理后信号
    Figure PCTCN2019110287-appb-100007
    S10步，输出第k时刻自适应滤波算法的输出信号o(k)；判断自适应滤波终止指示变量：若自适应滤波终止指示变量为真，则结束自适应滤波算法，否则跳至S2步计算下一时刻自适应滤波算法的输出。
  3. 根据权利要求2所述的用于采集体音信号的双麦克风自适应滤波算法，其特征在于：所述S1步中，滤波器阶数M的取值范围为：M∈[10,200]。
  4. 根据权利要求2所述的用于采集体音信号的双麦克风自适应滤波算法，其特征在于：所述S4步中，高通滤波处理采用如下两种方案之一：
    一、采用脉冲传递函数为G HP(z)的高通滤波处理器，脉冲传递函数G HP(z)的截止频率f HPc的取值范围为：500～1200Hz；
    二、采用由m HP个一阶预加重环节1-α jz -1，j＝1，...，m HP，α j∈[0.9,1)串联而构成的预加重高通滤波器。
  5. 根据权利要求4所述的用于采集体音信号的双麦克风自适应滤波算法，其特征在于：所述第一种方案中，S9步第一次低通滤波处理采用的低通滤波器的脉冲传递函数G 1LP(z)＝1/G HP(z)。
  6. 根据权利要求2所述的用于采集体音信号的双麦克风自适应滤波算法，其特征在于：所述S8步中，调节因子η的取值范围为：η∈[0.1,1]。
  7. 根据权利要求2所述的用于采集体音信号的双麦克风自适应滤波算法，其特征在于：所述S10步中，输出第k时刻自适应滤波算法的输出信号o(k)，是指：采用如下两种方式之一：
    一、将第一次低通滤波处理后信号
    Figure PCTCN2019110287-appb-100008
    作为第k时刻自适应滤波算法的输出信号o(k)来输出；
    二、对第一次低通滤波处理后信号
    Figure PCTCN2019110287-appb-100009
    作第二次低通滤波处理以进一步抑制环境噪音干扰，以第二次低通滤波处理后信号作为第k时刻自适应滤波算法的输出信号o(k)来输出。
  8. 根据权利要求7所述的用于采集体音信号的双麦克风自适应滤波算法，其特征在于：所述第二种方式中，第二次低通滤波处理采用脉冲传递函数为G 2LP(z)的低通滤波器，脉冲传递函数G 2LP(z)的截止频率f LPc的取值范围为：1200～1600Hz。
  9. 一种权利要求1至8中任一项所述的用于采集体音信号的双麦克风自适应滤波算法的应用，其特征在于：应用于电子听诊设备和/或电子穿戴设备，将自适应滤波算法输出的体音信号作为电子听诊设备和/或电子穿戴设备的输出信号。
PCT/CN2019/110287 2018-12-06 2019-10-10 用于采集体音信号的双麦克风自适应滤波算法及应用 WO2020114071A1 (zh)

Priority Applications (2)

Application Number Priority Date Filing Date Title
SG11202103089YA SG11202103089YA (en) 2018-12-06 2019-10-10 A Dual-Microphone Adaptive Filtering Algorithm for Collecting Body Sound Signals and Application Thereof
US17/282,761 US11735200B2 (en) 2018-12-06 2019-10-10 Dual-microphone adaptive filtering algorithm for collecting body sound signals and application thereof

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN201811485004.8A CN109545239B (zh) 2018-12-06 2018-12-06 用于采集体音信号的双麦克风自适应滤波算法及应用
CN201811485004.8 2018-12-06

Publications (1)

Publication Number Publication Date
WO2020114071A1 true WO2020114071A1 (zh) 2020-06-11

Family

ID=65854000

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2019/110287 WO2020114071A1 (zh) 2018-12-06 2019-10-10 用于采集体音信号的双麦克风自适应滤波算法及应用

Country Status (4)

Country Link
US (1) US11735200B2 (zh)
CN (1) CN109545239B (zh)
SG (1) SG11202103089YA (zh)
WO (1) WO2020114071A1 (zh)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN112754516A (zh) * 2021-02-07 2021-05-07 河南省肿瘤医院 一种智能化肠鸣音定位及采集装置

Families Citing this family (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN109545239B (zh) * 2018-12-06 2021-11-05 华南理工大学 用于采集体音信号的双麦克风自适应滤波算法及应用
TWI769449B (zh) 2020-04-21 2022-07-01 廣達電腦股份有限公司 濾波系統及濾波方法
CN112190280B (zh) * 2020-10-13 2022-10-14 苏州美糯爱医疗科技有限公司 一种电子听诊器背景音干扰实时自动抵消方法
CN112562701B (zh) * 2020-11-16 2023-03-28 华南理工大学 心音信号双通道自适应降噪算法、装置、介质及设备

Citations (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20100014723A1 (en) * 2008-07-15 2010-01-21 Nellcor Puritan Bennett Ireland Signal processing systems and methods using multiple signals
CN101763858A (zh) * 2009-10-19 2010-06-30 瑞声声学科技(深圳)有限公司 双麦克风信号处理方法
WO2013055394A1 (en) * 2011-10-14 2013-04-18 Advanced Fuel Research, Inc. Laser stethoscope
CN104581516A (zh) * 2013-10-15 2015-04-29 清华大学 一种医学声信号的双麦克风消噪方法及装置
CN106510751A (zh) * 2016-09-28 2017-03-22 华南师范大学 多功能电子听诊器及听诊系统
CN108574892A (zh) * 2018-06-29 2018-09-25 歌尔科技有限公司 拾音器、拾音设备及消除噪音的方法
CN109545239A (zh) * 2018-12-06 2019-03-29 华南理工大学 用于采集体音信号的双麦克风自适应滤波算法及应用

Family Cites Families (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US7747316B2 (en) * 2005-01-27 2010-06-29 The Board Of Trustees Of The University Of Illinois Blind adaptive filter extraction of fetal electrocardiogram signal estimate
CN102347027A (zh) * 2011-07-07 2012-02-08 瑞声声学科技(深圳)有限公司 双麦克风语音增强装置及其语音增强方法
US9848848B2 (en) * 2014-06-30 2017-12-26 The Johns Hopkins University Lung sound denoising stethoscope, algorithm, and related methods
EP3367908B1 (en) * 2015-10-30 2020-09-23 The Johns Hopkins University Programmable electronic stethoscope devices, algorithms, systems, and methods
CN106691376B (zh) * 2016-11-30 2019-05-28 深圳市科曼医疗设备有限公司 心电信号自适应滤波方法及装置
CN106782593B (zh) * 2017-02-27 2019-10-25 重庆邮电大学 一种用于声学回声消除的多带结构自适应滤波器切换方法

Patent Citations (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20100014723A1 (en) * 2008-07-15 2010-01-21 Nellcor Puritan Bennett Ireland Signal processing systems and methods using multiple signals
CN101763858A (zh) * 2009-10-19 2010-06-30 瑞声声学科技(深圳)有限公司 双麦克风信号处理方法
WO2013055394A1 (en) * 2011-10-14 2013-04-18 Advanced Fuel Research, Inc. Laser stethoscope
CN104581516A (zh) * 2013-10-15 2015-04-29 清华大学 一种医学声信号的双麦克风消噪方法及装置
CN106510751A (zh) * 2016-09-28 2017-03-22 华南师范大学 多功能电子听诊器及听诊系统
CN108574892A (zh) * 2018-06-29 2018-09-25 歌尔科技有限公司 拾音器、拾音设备及消除噪音的方法
CN109545239A (zh) * 2018-12-06 2019-03-29 华南理工大学 用于采集体音信号的双麦克风自适应滤波算法及应用

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN112754516A (zh) * 2021-02-07 2021-05-07 河南省肿瘤医院 一种智能化肠鸣音定位及采集装置

Also Published As

Publication number Publication date
US20220005491A1 (en) 2022-01-06
SG11202103089YA (en) 2021-04-29
US11735200B2 (en) 2023-08-22
CN109545239A (zh) 2019-03-29
CN109545239B (zh) 2021-11-05

Similar Documents

Publication Publication Date Title
WO2020114071A1 (zh) 用于采集体音信号的双麦克风自适应滤波算法及应用
US5610987A (en) Active noise control stethoscope
JP5345852B2 (ja) 生物音響センサ
EP2202998B1 (en) A device for and a method of processing audio data
JP3626198B2 (ja) 電子聴診器
AU2018266253B2 (en) System and method for determining cardiac rhythm and/or respiratory rate
Jatupaiboon et al. Electronic stethoscope prototype with adaptive noise cancellation
US7006638B1 (en) Electronic stethoscope
CN108600894A (zh) 一种耳机自适应有源噪声控制系统及方法
JP2009517127A (ja) 片持ち梁型生体音響センサー及びそのセンサーを使用した方法
McLane et al. Design and comparative performance of a robust lung auscultation system for noisy clinical settings
TWI827009B (zh) 聽覺補償裝置及聽覺補償方法
CN105356861B (zh) 一种有源降噪的方法和系统
Potdar et al. Design and Analysis of a PCG Acquisition and Denoising Circuit
CN115024719A (zh) 一种基于听力仿生的耳鸣检测方法及装置
Wang Multi-sensor adaptive heart and lung sound extraction
CN207369229U (zh) 一种耳机
Pasha et al. Multi-channel electronic stethoscope for enhanced cardiac auscultation using beamforming and equalisation techniques
Rong et al. Wearable Electro-Phonocardiography Device for Cardiovascular Disease Monitoring
CN107172516B (zh) 一种耳机及心率检测方法
CN111150422A (zh) 智能听诊系统及其听诊方法
WO2023077252A1 (zh) 一种基于FxLMS结构的主动降噪系统、方法及设备
CN113990339A (zh) 声音信号处理方法、装置、系统、设备和存储介质
US20230154449A1 (en) Method, device, headphones and computer program for actively suppressing interfering noise
US20230419981A1 (en) Audio signal processing method and system for correcting a spectral shape of a voice signal measured by a sensor in an ear canal of a user

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 19892740

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

32PN Ep: public notification in the ep bulletin as address of the adressee cannot be established

Free format text: NOTING OF LOSS OF RIGHTS PURSUANT TO RULE 112(1) EPC (EPO FORM 1205A DATED 15/09/2021)

122 Ep: pct application non-entry in european phase

Ref document number: 19892740

Country of ref document: EP

Kind code of ref document: A1