阅读说明：本技术 一种基于低秩张量学习的多通道心肺音异常识别系统与装置 (Multichannel cardiopulmonary sound abnormity identification system and device based on low-rank tensor learning ) 是由 邱育宁 谢胜利 谢侃 杨其宇 吕俊 周郭许 王艳娇 陈林楷 于 2020-06-24 设计创作，主要内容包括：本发明公开了一种基于低秩张量学习的多通道心肺音异常识别系统与装置,包括：通过3个拾音器组成的拾音器整列采集人体心肺音信号,利用增益调节器对心肺音信号进行放大后,通过滤波器,得到高信噪比的心肺音信号,通过ADC模块转换成数字信号传输至单片机；单片机对于3个通道采集的混合心肺音信号分别进行短时傅里叶变换处理,其中短时傅里叶变换采用的窗函数为汉明窗,对3通道信号进行短时傅里叶变换后,获得3个时频谱；根据采集的心肺音张量数据以及给定的标签,训练低秩张量分类模型,获得预训练的学习参数；当给定的新采集的听诊数据时,使用分类模型预测出患者的心肺音数据是否为异常。本发明能减少学习参数,能实现小样本心肺音异常识别任务。(The invention discloses a multichannel cardiopulmonary sound abnormity identification system and device based on low-rank tensor learning, which comprises the following steps: the method comprises the steps that a sound pick-up composed of 3 sound pick-ups is used for collecting heart and lung sound signals of a human body in an arraying mode, the heart and lung sound signals are amplified by a gain adjuster, then the heart and lung sound signals with high signal-to-noise ratio are obtained through a filter, and the heart and lung sound signals are converted into digital signals through an ADC (analog to digital converter) module and transmitted to a single chip microcomputer; the single chip microcomputer respectively carries out short-time Fourier transform processing on the mixed heart and lung sound signals collected by the 3 channels, wherein a window function adopted by the short-time Fourier transform is a Hamming window, and 3 time spectrums are obtained after the short-time Fourier transform is carried out on the 3 channels of signals; training a low-rank tensor classification model according to the acquired cardiopulmonary sound tensor data and a given label to obtain pre-training learning parameters; given newly acquired auscultation data, a classification model is used to predict whether the patient's cardiopulmonary sound data is abnormal. The invention can reduce learning parameters and realize the task of identifying the small sample cardiopulmonary sound abnormity.)

1. A multichannel cardiopulmonary sound abnormity identification system based on low rank tensor learning is characterized by comprising the following components:

the method comprises the steps that a sound pick-up composed of 3 sound pick-ups is used for collecting heart and lung sound signals of a human body in an arraying mode, the heart and lung sound signals are amplified by a gain adjuster, then the heart and lung sound signals with high signal-to-noise ratio are obtained through a filter, and the heart and lung sound signals are converted into digital signals through an ADC (analog to digital converter) module and transmitted to a single chip microcomputer;

the single chip microcomputer respectively carries out short-time Fourier transform processing on the mixed heart and lung sound signals collected by the 3 channels, wherein a window function adopted by the short-time Fourier transform is a Hamming window, 3 channel signals are subjected to short-time Fourier transform one by one to obtain 3 time frequency spectrums, and the 3 time frequency spectrums are expressed into 1 three-dimensional tensor with the size of 3T F, wherein T represents collection time length, and F represents frequency;

training a low-rank tensor classification model according to the acquired cardiopulmonary sound tensor data and a given label to obtain pre-training learning parameters; when newly acquired auscultation data is given, predicting whether the cardiopulmonary sound data of the patient is abnormal by using a classification model;

the single chip microcomputer outputs the prediction result to the display module to remind the user.

2. The system for recognizing the abnormal heart-lung sounds based on the low-rank tensor learning as claimed in claim 1, wherein the length of the hamming window is set to 100-;

3. the system for recognizing abnormality of cardiopulmonary sound based on low rank tensor learning of claim 1, wherein the three-dimensional tensor data is expressed as follows:

the low rank constraint of the 3-dimensional tensor can be expressed as a weighted sum of the kernel norms spread for 3 modalities:

wherein | · | purple_*Is a matrix kernel norm, W_[k]Is tensorThe modulo-k matrix of (c) is expanded.

4. The system for identifying the abnormal heart-lung sounds based on the low-rank tensor learning as claimed in claim 1, wherein the model construction comprises the following processes:

the heart and lung sound signals are divided into 4 categories, namely 1) the heart sound and lung sound are normal, 2) the heart sound and lung sound are normal and abnormal, 3) the heart sound and lung sound are abnormal and 4) the heart sound and lung sound are abnormal. Through the divided data sets, the parameters of 4 categories are learned by respectively using the proposed multichannel cardiopulmonary sound abnormality identification method based on tensor low-rank learningThe objective function is as follows:

wherein C is_rNumber of cardiopulmonary sound training samples for the r-th category_， And b^rParameters and deviations, respectively, of the r-th class_， Andthe kth auxiliary variable and the dual variable of the category r respectively;

3) solving an objective function

Alternately updating each variable of the objective function by adopting an alternate multiplier updating method

a) Updating auxiliary variablesWherein k is 1,2, 3:

whereinA singular value contraction operator;

b) updating learning parameters

c) Updating dual variablesWherein k is 1,2, 3:

d) checking whether the following convergence conditions are reached, and if so, stopping iterative learning:

whereinLearning parameters for a previous iteration;

by alternately updating the variables, the target function can reach a local minimum value point, so that the learning parameter of each category is obtainedAnd b^r。

5. The system for identifying abnormal heart-lung sounds based on low rank tensor learning as claimed in claim 4, wherein the classification comprises the following processes: to test the sample4 learned multi-channel cardiopulmonary sound abnormality recognition systems are input, and then the outputs of the 4 learners are calculated respectively. Comparing the outputs of 4 different recognition systems, selecting a system corresponding to the maximum output value to determine which category the input cardiopulmonary sound signal belongs to, and adopting the following formula:

representing predictionsThe category (2).

6. A multi-channel cardiopulmonary sound abnormity identification device is characterized by comprising a data processor, a sound pick-up, a gain adjuster, a filter, an analog-to-digital converter and a display module; the sound pick-up, the gain adjuster, the filter and the analog-to-digital converter are sequentially connected, and the analog-to-digital converter and the display module are respectively connected with the data processor.

7. The multi-channel cardiopulmonary sound abnormality recognition device of claim 6, wherein the sound pickup is one of an electret microphone, a microphone or other sensors that convert sound signals into electrical signals;

8. the device for recognizing the abnormal heart-lung sounds according to claim 6, further comprising a multi-channel sound collection cavity, wherein the multi-channel sound collection cavity is arranged on a sound pickup, and a sound pickup is arranged in each channel.

9. The apparatus as claimed in claim 6, wherein the gain adjuster is an operational amplifier, and the operational amplifier is of type TLC 2274.

10. The device as claimed in claim 6, wherein the analog-to-digital converter is ADS 1115.

Technical Field

The invention relates to the field of intelligent electronic auscultation, in particular to a low-rank tensor learning-based multi-channel cardiopulmonary sound abnormality identification system and device.

Background

In recent years, intelligent analysis of cardiopulmonary sounds has progressed to present time-frequency analysis in the time domain, frequency domain, and power spectrum. Time-frequency spectral analysis of cardiopulmonary sounds has become a currently effective and popular method. In order to be able to efficiently identify the presence of abnormal Heart-lung sounds based on their time-frequency spectrum, research scientists have proposed a number of methods including support vector machine (C.Sowmiya and P.Sumitra, "Analytical study of Heart discrete diagnosis classification Techniques,"2017 IEEE International Conference Intelligent technique in Control, Optimization and Signal Processing (INCOS), Srivillitus, 2017, pp.1-5), Linear discriminant analysis (P.mayorga, J.Valdez, C.DruzzlisandedV.Zeljkovic, "Heart and lubricating base analysis, classification 2016," Global Engineering analysis/amplification/analysis/GME 3531, PAid-E75E, PAmdiei, PAE 752016), methods such as "Classification of heart sound signalling multiple services", "Applied Sciences 8.12(2018): 2344", and the like, but these methods have two more serious problems. On one hand, after the time frequency spectrum is simply vectorized, the method is further used for carrying out abnormity detection on the time frequency spectrum, and the simple vectorization mode of the time frequency spectrum can destroy the time and space structure of data, so that a better learning effect cannot be achieved; on the other hand, the large-scale learning parameters of the classifier need to provide a large number of training samples, and are not in accordance with the practical requirements.

Disclosure of Invention

The invention aims to overcome the defects of the prior art and provides a multichannel cardiopulmonary sound abnormality identification system based on low-rank tensor learning, which comprises the following steps:

the method comprises the steps that a sound pick-up composed of 3 sound pick-ups is used for collecting heart and lung sound signals of a human body in an arraying mode, the heart and lung sound signals are amplified by a gain adjuster, then the heart and lung sound signals with high signal-to-noise ratio are obtained through a filter, and the heart and lung sound signals are converted into digital signals through an ADC (analog to digital converter) module and transmitted to a single chip microcomputer;

the single chip microcomputer respectively carries out short-time Fourier transform processing on the mixed heart and lung sound signals collected by the 3 channels, wherein a window function adopted by the short-time Fourier transform is a Hamming window, 3 channel signals are subjected to short-time Fourier transform one by one to obtain 3 time frequency spectrums, and the 3 time frequency spectrums are expressed into 1 three-dimensional tensor with the size of 3T F, wherein T represents collection time length, and F represents frequency;

training a low-rank tensor classification model according to the acquired cardiopulmonary sound tensor data and a given label to obtain pre-training learning parameters; when newly acquired auscultation data is given, predicting whether the cardiopulmonary sound data of the patient is abnormal by using a classification model;

the single chip microcomputer outputs the prediction result to the display module to remind the user.

Further, the length of the Hamming window is set to 100-;

further, the three-dimensional tensor data is expressed by the following formula:

the low rank constraint of the 3-dimensional tensor can be expressed as a weighted sum of the kernel norms spread for 3 modalities:

where | · | > is the matrix kernel norm, W_[κ]Is tensorThe modulo-k matrix of (c) is expanded.

Further, the model construction comprises the following processes:

the heart and lung sound signals are divided into 4 categories, namely 1) the heart sound and lung sound are normal, 2) the heart sound and lung sound are normal and abnormal, 3) the heart sound and lung sound are abnormal and 4) the heart sound and lung sound are abnormal. Through the divided data sets, the parameters of 4 categories are learned by respectively using the proposed multichannel cardiopulmonary sound abnormality identification method based on tensor low-rank learningThe objective function is as follows:

wherein C is_rThe number of the cardiopulmonary sound training samples of the r-th category,and b^rRespectively the parameter and the deviation of the r-th class,andthe kth auxiliary variable and the dual variable of the category r respectively;

3) solving an objective function

Alternately updating each variable of the objective function by adopting an alternate multiplier updating method

a) Updating auxiliary variablesWherein k is 1,2, 3:

whereinA singular value contraction operator;

b) updating learning parameters

c) Updating dual variablesWherein k is 1,2, 3:

d) checking whether the following convergence conditions are reached, and if so, stopping iterative learning:

whereinLearning parameters for a previous iteration;

by alternately updating the above variables, the target function can be made to reach a local minimum point, thereby obtainingLearning parameters for each classAnd b^T。

Further, the classification includes the following processes: to test the sample4 learned multi-channel cardiopulmonary sound abnormality recognition systems are input, and then the outputs of the 4 learners are calculated respectively. Comparing the outputs of 4 different recognition systems, selecting a system corresponding to the maximum output value to determine which category the input cardiopulmonary sound signal belongs to, and adopting the following formula:

representing predictionsThe category (2).

A multi-channel cardiopulmonary sound abnormity identification device comprises a data processor, a sound pickup, a gain adjuster, a filter, an analog-to-digital converter and a display module; the sound pick-up, the gain adjuster, the filter and the analog-to-digital converter are sequentially connected, and the analog-to-digital converter and the display module are respectively connected with the data processor.

Preferably, the sound pick-up adopts one of an electret microphone, a microphone or other sensors for converting sound signals into electric signals;

preferably, still include multichannel collection sound cavity, multichannel collection sound cavity set up on the adapter, set up an adapter in each passageway.

Preferably, the gain adjuster is an operational amplifier, and the operational amplifier is of a type TLC 2274.

Preferably, the analog-to-digital converter is of the type ADS 1115.

The invention has the beneficial effects that: the heart and lung sounds are not required to be manually extracted, manual intervention is reduced, and the identification accuracy can be improved; the method can reduce learning parameters and realize the task of recognizing the small sample cardiopulmonary sound abnormity; the invention can effectively reserve the spatial information in the time frequency spectrums of a plurality of channels and effectively improve the identification accuracy.

Drawings

Fig. 1 is a schematic diagram of a multichannel cardiopulmonary sound abnormality identification system based on low rank tensor learning;

FIG. 2 is a schematic diagram of a multi-channel cardiopulmonary sound abnormality recognition apparatus;

FIG. 3 is a schematic diagram of a single heart-lung sound signal acquisition channel.

Detailed Description

The technical solutions of the present invention are further described in detail below with reference to the accompanying drawings, but the scope of the present invention is not limited to the following.

As shown in fig. 1, in step 1, a sound pickup composed of 3 sound pickups is used to collect heart and lung sound signals of a human body in an array, the heart and lung sound signals are amplified by a gain adjuster, then the heart and lung sound signals with high signal-to-noise ratio are obtained by a filter, and are converted into digital signals by an ADC module and transmitted to a single chip microcomputer;

and 2, respectively carrying out short-time Fourier transform processing on the mixed heart-lung sound signals acquired by the 3 channels by the singlechip, wherein the window function adopted by the short-time Fourier transform is a Hamming window, the length of the Hamming window is set to be 100-128, and the compensation of window movement is set to be 50-64. After short-time Fourier transform is carried out on the 3-channel signals one by one, 3 time frequency spectrums can be obtained, so that the obtained 3 time frequency spectrums can be naturally expressed into 1 three-dimensional tensor with the size of 3T F, wherein T represents the acquisition time length, F represents the frequency, and the frequency range of 20Hz-1000Hz is selected as the frequency range of heart sound is 20Hz-600Hz and the frequency range of lung sound is 60Hz-1000 Hz; training a low-rank tensor classification model according to the acquired cardiopulmonary sound tensor data and a given label to obtain pre-training learning parameters; when newly acquired auscultation data is given, predicting whether the cardiopulmonary sound data of the patient is abnormal by using a classification model;

step 3, the single chip microcomputer outputs the prediction result of the step 2 to a display module to remind a user;

the method specifically comprises the following steps: the sound pick-up can comprise an electret microphone, a microphone or other sensors for converting sound signals into electric signals, and sound vibration signals are converted into the electric signals through the sound pick-up, wherein a sound pick-up array is adopted to be matched with a multi-channel sound collecting cavity, as shown in a figure X, and one stethoscope head comprises 3 sound cavities; as shown in fig. X, U1 is an operational amplifier of type TLC2274, the operational amplifier cooperates with a resistance-capacitance network to form a two-stage inverting amplifier and a band-pass filter to amplify and filter the collected cardiopulmonary sounds, wherein the gain of the inverting amplifier is 40dB, and the upper and lower cut-off frequencies of the band-pass filter are 100Hz and 3kHz, respectively; in a specific implementation process, the sound pickup, the gain adjuster and the filter are provided with 3 channels, and after cardiopulmonary sound signals of the 3 channels are obtained, the signals are transmitted to the ADC module; in the specific implementation process, the ADC is selected to be the ADS1115, the ADS1115 is provided with 4 ADC acquisition channels, and the analog cardiopulmonary sound signals can be converted into digital signals and transmitted to the single chip microcomputer through the IIC interface; in the specific implementation process, the single chip microcomputer is STM32H743ZI, the algorithm program runs in the single chip microcomputer, and the result is output to the display module after the prediction structure is obtained; in the specific implementation process, the display module is a 0.91-inch OLED12832 display screen and is connected with the single chip microcomputer through an IIC interface.

Tensor data representing original time-frequency spectrum of multiple channels into three-dimension

1) Is to the parameter tensorLow rank constraint is performed to learn low rank discrimination information useful for the cardiopulmonary sound recognition classifier.

The low rank constraint of the 3-dimensional tensor can be expressed as a weighted sum of the kernel norms spread for 3 modalities:

where | · | > is the matrix kernel norm, W_[κ]Is tensorThe modulo-k matrix of (c) is expanded.

2) Constructing an objective function

To identify abnormal heart-lung sound signals, we classified the heart-lung sound signals into 4 categories, which are 1) normal heart sound and lung sound, 2) normal heart sound and lung sound, 3) abnormal heart sound and lung sound, and 4) abnormal heart sound and lung sound. Through the divided data sets, the parameters of 4 categories are learned by respectively using the proposed multichannel cardiopulmonary sound abnormality identification method based on tensor low-rank learningThe objective function is as follows:

wherein C is_rThe number of the cardiopulmonary sound training samples of the r-th category,and b^rRespectively the parameter and the deviation of the r-th class,andthe kth auxiliary variable and the dual variable of the class r, respectively.

3) Solving an objective function

Alternately updating each variable of the objective function by adopting an alternate multiplier updating method

e) Update assistanceVariables ofWherein k is 1,2, 3:

whereinA singular value contraction operator;

f) updating learning parameters

g) Updating dual variablesWherein k is 1,2, 3:

h) checking whether the following convergence conditions are reached, and if so, stopping iterative learning:

whereinThe learning parameters of the previous iteration.

By alternately updating the variables, the target function can reach a local minimum value point, so that the learning parameter of each category is obtainedAnd b^r。

Using learned parametersAnd b^rClassifying the cardiopulmonary sound test samples, wherein the classification steps are as follows:

based on the above learned parameters, test samples are tested4 learned multi-channel cardiopulmonary sound abnormality recognition systems are input, and then the outputs of the 4 learners are calculated respectively. And comparing the outputs of the 4 different recognition systems, and selecting the system corresponding to the maximum output value to determine the category of the input cardiopulmonary sound signal. Namely, it is expressed in the following form:

representing predictionsThe category (2).

The foregoing is illustrative of the preferred embodiments of this invention, and it is to be understood that the invention is not limited to the precise form disclosed herein and that various other combinations, modifications, and environments may be resorted to, falling within the scope of the concept as disclosed herein, either as described above or as apparent to those skilled in the relevant art. And that modifications and variations may be effected by those skilled in the art without departing from the spirit and scope of the invention as defined by the appended claims.