Bowel sound detection system

Abstract

The invention relates to a borborygmus detection system, belongs to the technical field of borborygmus detection, and solves the problem of inaccurate borborygmus recognition in the prior art. The system comprises a signal acquisition device which comprises a laser, a distributed optical fiber sensor and a collector; the laser is used for periodically emitting laser to the distributed optical fiber sensor; the collector is used for collecting a signal returned by the optical fiber sensor and sending the signal to the borborygmus resolving module; the borborygmus resolving module is used for resolving the acquired signal based on the optical frequency domain reflection, judging whether borborygmus occurs or not, and if so, determining the generation position of the borborygmus and extracting a borborygmus signal; and the borborygmus classification module is used for obtaining the borborygmus type of the individual based on a trained borborygmus classification model according to the generation position of the borborygmus and the borborygmus signal. Therefore, the borborygmus can be accurately identified.

Description

Technical Field

[0001] The present invention relates to the technical field of bowel sound detection, and in particular to a bowel sound detection system. Background Art

[0002] Bowel sounds are the sounds produced by the friction between intestinal contents (solid, gas or liquid) and the intestines, or between intestinal contents, under the premise of intestinal peristalsis. They can reflect the state of intestinal movement and are related to clinical applications such as the physiological state of the human body and disease diagnosis.

[0003] At present, clinical practice mainly relies on manual auscultation of bowel sounds, which has defects such as strong subjectivity, few evaluation parameters, high missed diagnosis rate, and data cannot be recorded and repeated. At the same time, the quality of auscultation is limited by the experience and ability of the auscultator. From the perspective of informatics, bowel sounds are a kind of vibration signal, which contains a wealth of disease-related information, such as main frequency, loudness, etc. This information is difficult for the human ear to recognize, or it cannot be directly recognized. Compared with other physiological sounds of the human body (such as heart sounds and respiratory sounds), bowel sounds have poor periodicity, low regularity, strong timeliness, and many influencing factors, which brings many difficulties to clinical application. Therefore, the accuracy and clinical value of manual auscultation of bowel sounds face great challenges.

[0004] Existing bowel sound collection equipment usually uses traditional acoustic sensors such as coil microphones, and fixes one or more sensors near the abdominal cavity through a belt to achieve bowel sound monitoring. The belt-type acoustic sensor array is in contact with the human body, and a large amount of friction noise will be generated during human activities, which will seriously interfere with the bowel sound signal. At the same time, the noise signal will greatly increase the amount of data and reduce the data quality, which will bring difficulties to the subsequent bowel sound identification. The combination of the belt-type sensor array and the traditional acoustic probe limits the number of sensors, making the time-delay acoustic positioning method less accurate; because the wavelength of the sound wave is much larger than the size of the abdominal cavity, and the reflection and superposition of the sound wave in the human tissue are more serious, the traditional method of locating the position of bowel sounds has poor accuracy. In addition, due to the differences in the human body and the movement during wearing, the initial position of the sensor will be inaccurate and displaced, resulting in the failure of bowel sound positioning and the degradation of data quality. Therefore, the current monitoring and identification of bowel sounds has many deficiencies in the basic stage of data collection, which greatly increases the difficulty of subsequent bowel sound recognition. Summary of the invention

[0005] In view of the above analysis, an embodiment of the present invention aims to provide a bowel sound detection system to solve the problem of inaccurate bowel sound recognition in the prior art.

[0006] In one aspect, an embodiment of the present invention provides a bowel sound detection system, the system comprising: The signal acquisition device comprises a laser, a distributed optical fiber sensor and a collector; the laser is used to periodically emit laser light to the distributed optical fiber sensor; the collector is used to collect the signal returned by the optical fiber sensor and send it to the bowel sound solution module; The bowel sound calculation module is used to calculate the collected signal based on optical frequency domain reflection to determine whether bowel sounds occur. If so, the location of the bowel sounds is determined and the bowel sound signals are extracted; The bowel sound classification module is used to obtain the individual bowel sound type based on the location of the bowel sound and the bowel sound signal based on the trained bowel sound classification model.

[0007] Based on the further improvement of the above technical solution, the distributed optical fiber sensor includes a test optical fiber, a reference optical fiber, a first coupler and a circulator; the first coupler is connected to the laser and is used to divide the light source into two light paths, one of which is injected into the reference optical fiber and the other is injected into the test optical fiber through the circulator; The collector includes a second coupler and a photodetector. The second coupler is connected to the reference optical fiber and the circulator, and is used to couple the signals returned by the reference optical fiber and the test optical fiber and send them to the photodetector; the photodetector converts the optical signal into an electrical signal and sends it to the bowel sound solution module.

[0008] Based on the further improvement of the above technical solution, the signal acquisition device also includes a cylindrical shell; there is an annular groove on the inner wall of the cylindrical shell; the test optical fiber of the distributed optical fiber sensor is wound on the inner wall of the cylindrical shell, and the test optical fiber of the distributed optical fiber sensor is embedded in the groove on the inner wall of the cylindrical shell.

[0009] Based on the further improvement of the above technical solution, the bowel sound solution module uses the following method to solve the collected signal to determine whether bowel sounds occur, and if so, determine the location of the bowel sounds: Acquire a reference signal; the reference signal has the same length as the acquisition signal; Whether bowel sounds occur is determined based on the similarity between the reference signal and the collected signal, and if so, the location of the bowel sounds is located based on the similarity.

[0010] Based on the further improvement of the above technical solution, the following method is used to determine whether bowel sounds occur based on the similarity between the reference signal and the collected signal: The reference signal and the acquisition signal are segmented respectively according to the emission period of the laser; The reference signal and each segmented signal of the acquired signal are converted into the range domain by fast Fourier transform; For each distance domain segment of the collected signal, the distance domain segment of the collected signal and the corresponding distance domain segment of the reference signal are respectively divided into a plurality of subdivided segments by using a sliding window; For each subdivision of the distance domain segmentation, the subdivision and the subdivision of the reference signal corresponding to the subdivision are respectively converted into time domain signals by using inverse Fourier transform; the cross-correlation diagram of the two groups of time domain signals is calculated to obtain the similarity of the subdivision; If there is a subdivision whose similarity is less than the first threshold in the distance segment, it is determined that bowel sounds have occurred.

[0011] Based on the further improvement of the above technical solution, the bowel sound classification model includes: A convolutional coding module is used to extract features from the bowel sound signal to obtain a feature frame sequence; the feature frame sequence is randomly masked and input into a BERT coding module; BERT encoding module, which is used to output the hidden representation of each feature frame based on the sequence of feature frames after random masking; A classification module is used to predict the type of bowel sounds based on the location of the bowel sounds and the hidden representation of each feature frame.

[0012] Based on the further improvement of the above technical solution, the trained bowel sound classification model is obtained in the following way: Acquire multiple bowel sound signals to construct a training sample set; A convolutional coding module and a BERT coding module are combined into a pre-training model; a clustering module is added to the pre-training model; the clustering module is used to cluster the signal frames of the bowel sound signal to obtain a label of each cluster type; the BERT coding module also outputs a prediction result of the cluster type of each feature frame; The pre-trained model is trained based on the label of the clustering type of each feature frame of the sample and the predicted clustering type calculation loss to obtain a trained pre-trained model; The parameters of the convolutional coding module are fixed, and the bowel sound classification model is fine-tuned based on the training sample set to obtain a trained bowel sound classification model.

[0013] Based on the further improvement of the above technical solution, the following formula is used to calculate the loss: ; in, represents the loss of the masked feature frame in the sample, represents the loss of the unmasked feature frame in the sample, Represents weight.

[0014] Based on the further improvement of the above technical solution, the loss of the masked feature frame is calculated using the following formula: ; in, represents the clustering type loss of the i-th masked feature frame, represents the distance loss of the i-th masked feature frame, Indicates the number of masked feature frames in the sample.

[0015] Based on the further improvement of the above technical solution, the distance loss of the i-th masked feature frame is calculated using the following formula: ; in, represents the hidden representation of the i-th masked feature frame, represents the hidden representation of the p-th feature frame, represents the hidden representation of the qth feature frame; the cluster type label of the pth feature frame is the same as the cluster type label of the i-th masked feature frame, and the cluster type label of the qth feature frame is different from the cluster type label of the i-th masked feature frame. Represents the margin parameters.

[0016] Compared with the prior art, the present invention collects signals in a non-contact manner through a signal acquisition device, judges and resolves bowel sounds through a bowel sound resolution module, thereby accurately resolving the occurrence location and sound signals, and predicts the type based on a trained bowel sound classification model through a bowel sound classification module, thereby accurately identifying bowel sounds.

[0017] In the present invention, the above-mentioned technical solutions can also be combined with each other to achieve more preferred combination solutions. Other features and advantages of the present invention will be described in the subsequent description, and some advantages can become obvious from the description, or can be understood by practicing the present invention. The purpose and other advantages of the present invention can be realized and obtained through the contents particularly pointed out in the description and the drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] The accompanying drawings are only used for the purpose of illustrating specific embodiments and are not to be considered as limiting the present invention. In the entire drawings, the same reference symbols represent the same components; Figure 1 is a block diagram of a bowel sound detection system according to an embodiment of the present invention; Figure 2 A partial structural schematic diagram of a signal acquisition device implemented in the present invention; Figure 3 A schematic diagram of the use of a signal acquisition device implemented in the present invention; Reference numerals: 1-the outer shell of the cylindrical shell, 2-the optical fiber, 3-the movable bed, 4-the sound insulation material, 5-the bell mouth, 6-the inner shell of the cylindrical shell, 7-the optical fiber section, 8-the shock absorbing chamber. DETAILED DESCRIPTION

[0019] The preferred embodiments of the present invention are described in detail below in conjunction with the accompanying drawings, wherein the accompanying drawings constitute a part of this application and are used together with the embodiments of the present invention to illustrate the principles of the present invention, but are not used to limit the scope of the present invention.

[0020] Optical fiber is a new type of communication medium. As a signal transmission carrier, its transmission distance ranges from a few meters to tens or even hundreds of kilometers. In many fields, it has gradually replaced traditional wireless communication technology. When light is transmitted in the optical fiber, photons collide with material molecules and produce scattering. Distributed optical fiber sensing realizes distributed optical fiber sensing by detecting the one-to-one correspondence between these scattered signals and external physical quantities.

[0021] Any position on the test optical fiber of a distributed optical fiber sensor can be used as a sensor, so the number of measurement points is much higher than the traditional sensor collection method. In addition, the distributed optical fiber sensor does not need to contact the individual being measured, so the positioning is more accurate and the collected signal quality is higher, thereby improving the accuracy of detection.

[0022] Based on this, a specific embodiment of the present invention discloses a bowel sound detection system, such as Figure 1 As shown, including: The signal acquisition device comprises a laser, a distributed optical fiber sensor and a collector; the laser is used to periodically emit laser light to the distributed optical fiber sensor; the collector is used to collect the signal returned by the optical fiber sensor and send it to the bowel sound solution module; The bowel sound calculation module is used to calculate the collected signal based on optical frequency domain reflection to determine whether bowel sounds occur. If so, the location of the bowel sounds is determined and the bowel sound signals are extracted; The bowel sound classification module is used to obtain the individual bowel sound type based on the location of the bowel sound and the bowel sound signal based on the trained bowel sound classification model.

[0023] Compared with the prior art, the bowel sound detection system provided in this embodiment collects signals in a non-contact manner through a signal acquisition device, and performs bowel sound judgment and interpretation through a bowel sound interpretation module, thereby accurately interpreting the occurrence location and sound signals, and performs type prediction based on a trained bowel sound classification model through a bowel sound classification module, thereby accurately identifying bowel sounds.

[0024] During implementation, the light source of the laser is a tunable laser source, and the laser emits a linear frequency sweep signal in each emission cycle.

[0025] The distributed optical fiber sensor comprises a test optical fiber, a reference optical fiber, a first coupler and a circulator; the first coupler is connected to a laser and is used to divide the light source into two optical paths, one of which is injected into the reference optical fiber and the other is injected into the test optical fiber through the circulator; The collector includes a second coupler and a photodetector. The second coupler is connected to the reference optical fiber and the circulator, and is used to couple the signals returned by the reference optical fiber and the test optical fiber and send them to the photodetector; the photodetector converts the optical signal into an electrical signal and sends it to the bowel sound solution module.

[0026] The light source is divided into two beams through the first coupler, one beam enters the reference optical fiber; the other beam enters the circulator through port 1 and is emitted through port 2 to enter the test optical fiber. Part of the light in the test optical fiber will be reflected back, and the reflected light enters the second coupler through port 3 of the circulator. The two light paths have different optical paths and optical frequencies, so a beat frequency signal is generated at the second coupler. The photodetector collects the beat frequency signal and converts it into an electrical signal, which is sent to the bowel sound resolution module.

[0027] During implementation, in order to facilitate signal collection, the signal collection device also includes a cylindrical shell; there is an annular groove on the inner wall of the cylindrical shell; the test optical fiber of the distributed optical fiber sensor is wound on the inner wall of the cylindrical shell, and the test optical fiber of the distributed optical fiber sensor is embedded in the groove on the inner wall of the cylindrical shell.

[0028] During implementation, annular grooves are arranged at equal intervals on the inner wall of the cylindrical shell.

[0029] When implementing, Figure 2 As shown, the cylindrical shell includes an outer shell and an inner wall, and the space between the outer shell and the inner wall is filled with sound insulation material. The test optical fiber is wrapped in a shock-absorbing cavity, thereby reducing the influence of the sensor on the outside world. In order to improve the detection sensitivity, the shock-absorbing cavity on the measuring surface of the optical fiber adopts a thin shell, or no shell, so that the optical fiber can detect physiological sound signals more accurately. By setting a groove on the inner wall, the test optical fiber is embedded in the groove, and the cross-section of the groove is in the shape of a trumpet, forming a trumpet mouth. The configuration of the trumpet mouth can enable the sensor to monitor a specific small area of ​​the abdomen in a targeted manner, thereby avoiding environmental interference and vibration interference from other parts (such as breathing, heartbeat, etc.), and focusing on the sound signal collection of the small area corresponding to the trumpet mouth, achieving physical noise reduction, and more suitable for receiving bowel sound signals.

[0030] During implementation, the individual to be tested lies on the mobile bed, and the mobile bed is pushed into the cylindrical shell for easy testing. Figure 3 The position corresponding to the abdomen on the bed is a hollow structure, so that the patient's back directly faces the lower touch sensor array to avoid affecting the transmission of bowel sound signals.

[0031] During implementation, the signal acquisition device collects signals for 2 to 4 minutes and sends the collected signals to the bowel sound interpretation module.

[0032] During implementation, after receiving the collected signal, the bowel sound calculation module uses the following method to calculate the collected signal to determine whether bowel sounds occur, and if so, determine the location of the bowel sounds: S21, obtaining a reference signal; the reference signal has the same length as the acquisition signal; S22. Determine whether bowel sounds occur based on the similarity between the reference signal and the collected signal, and if so, locate the location of the bowel sounds based on the similarity.

[0033] During implementation, the reference signal is a signal collected by the signal collection device when there is no sound. The reference signal has the same length as the signal collected during detection. Whether bowel sounds occur is determined based on the similarity between the reference signal and the collected signal.

[0034] Specifically, the following method is used to determine whether bowel sounds occur based on the similarity between the reference signal and the collected signal: S221, segmenting the reference signal and the acquisition signal respectively according to the emission period of the laser; S222, converting the reference signal and each segmented signal of the collected signal into a distance domain by fast Fourier transform; S223, for each range domain segment of the collected signal, using a sliding window to divide the range domain segment of the collected signal and the corresponding range domain segment of the reference signal into a plurality of sub-segments; S224, for each subdivision of the distance domain segmentation, using inverse Fourier transform to convert the subdivision and the subdivision of the reference signal corresponding to the subdivision into time domain signals respectively; calculating the cross-correlation diagram of the two groups of time domain signals to obtain the similarity of the subdivision; S225: If there is a sub-segment with a similarity less than a first threshold in the distance segment, it is determined that bowel sounds have occurred.

[0035] During implementation, firstly, the reference signal and the collected signal are segmented according to the emission period of the laser, and each segment corresponds to a signal collected by laser detection. The segmentation of the reference signal corresponds to the segmentation of the collected signal one by one.

[0036] For each group of segments (the i-th segment of the acquired signal and the i-th segment of the reference signal ), the reference signal and each segmented signal of the acquisition signal are converted to the frequency domain by fast Fourier transform, and the frequency domain signal is multiplied by a coefficient and converted into the distance domain signal of the test fiber in proportion, so as to obtain the corresponding distance domain segmented signal. and .

[0037] Use a sliding window to segment the i-th distance domain and The sliding windows are divided into a plurality of subdivision segments. During implementation, the sliding step length of the sliding window may be smaller than the window length, that is, two adjacent windows may overlap. The subdivisions and The subdivisions are one-to-one corresponding.

[0038] Each subdivided segment corresponds to a corresponding position of the test optical fiber.

[0039] For each subdivision, the jth subdivision and , use inverse Fourier transform to convert it into a time domain signal, and calculate the similarity of the two time domain signals. If there is a subdivision segment with a similarity less than a preset threshold in the i-th distance segment, it is determined that bowel sounds occur in the i-th segment.

[0040] During implementation, the cross-correlation diagram of two groups of time domain signals is calculated in the following way to obtain the similarity of the subdivision segment: The weighted cross-correlation function is used to perform cross-correlation processing on two groups of time domain signals to obtain a cross-correlation diagram; According to the formula Get the similarity of the subdivision segments, where represents the number of points in the cross-correlation graph whose amplitude exceeds the second threshold, Indicates the total number of points in the cross-correlation graph. During implementation, the first threshold and the second threshold are set according to the requirements of detection sensitivity and false alarm rate.

[0041] If there is a subdivision with a similarity less than the first threshold in the i-th distance segment, it is considered that bowel sounds occur at the collection time corresponding to the i-th distance segment, and the position corresponding to the subdivision with the smallest similarity is the position where the bowel sounds occur, thereby determining the location of the bowel sounds.

[0042] During implementation, the collected signal can be demodulated by a phase demodulation algorithm to obtain a bowel sound signal.

[0043] After obtaining the bowel sound signal, the occurrence location and the bowel sound signal are input into a trained bowel sound classification model to obtain the individual's bowel sound type.

[0044] It should be noted that the types of bowel sounds include hyperactive, active, normal, weakened, and absent.

[0045] Specifically, the bowel sound classification model includes: A convolutional coding module is used to extract features from the bowel sound signal to obtain a feature frame sequence; the feature frame sequence is randomly masked and input into a BERT coding module; BERT encoding module, which is used to output the hidden representation of each feature frame based on the sequence of feature frames after random masking; A classification module is used to predict the type of bowel sounds based on the location of the bowel sounds and the hidden representation of each feature frame.

[0046] During implementation, the classification module can use an existing classifier.

[0047] During implementation, the trained bowel sound classification model is obtained in the following manner: Acquire multiple bowel sound signals to construct a training sample set; A convolutional coding module and a BERT coding module are combined into a pre-training model; a clustering module is added to the pre-training model; the clustering module is used to cluster the signal frames of the bowel sound signal to obtain a label of each cluster type; the BERT coding module also outputs a prediction result of the cluster type of each feature frame; The pre-trained model is trained based on the label of the clustering type of each feature frame of the sample and the predicted clustering type calculation loss to obtain a trained pre-trained model; The parameters of the convolutional coding module are fixed, and the bowel sound classification model is fine-tuned based on the training sample set to obtain a trained bowel sound classification model.

[0048] During implementation, multiple bowel sound signals are obtained to construct a training sample set. Due to the low efficiency of manual labeling, only some bowel sound signals are required to have corresponding bowel sound type labels. By pre-training on unlabeled samples and then fine-tuning on labeled samples, the performance and efficiency of the model can be improved.

[0049] During implementation, the sound signal is a continuous time series signal. In order to extract signal features more deeply and improve the accuracy of detection, the bowel sound classification model extracts features from the bowel sound signal through the convolutional coding module and divides the bowel sound signal into multiple continuous feature frames. For example, feature extraction is performed on the bowel sound signal with a sampling rate of 16kHz to obtain a feature sequence with a frame rate of 20ms. During implementation, the convolutional coding module can use an existing convolutional neural network module.

[0050] During implementation, the original bowel sound signal is frame segmented according to the frame rate obtained by the convolutional coding module, that is, the length of each feature frame, to obtain the signal frame of the original bowel sound signal, and the signal frame and the feature frame correspond one to one. Pre-training is based on unlabeled samples. In order to train the model, a clustering module is added to the pre-training model to cluster each signal frame and add clustering labels, thereby training the model.

[0051] During implementation, the acoustic features (such as MFCC) of the original signal frame can be used as the input of the clustering module. After clustering, the centroid or center point of each cluster type is used as the label of the cluster type.

[0052] During implementation, the obtained feature frame sequence is randomly masked and input into the BERT encoding module. The BERT encoding module predicts and outputs the hidden representation of each feature frame based on the feature frame sequence after random masking, and also outputs the prediction result of the clustering type of each feature frame.

[0053] During implementation, the BERT encoding module adopts the existing BERT encoder structure and projects the hidden representation corresponding to each feature frame output by the BERT encoder to the space corresponding to the cluster label through the projection matrix, that is, the cluster type prediction result corresponding to the feature frame is obtained, and the loss is calculated based on the cluster type label and prediction result of the feature frame, and the pre-trained model parameters are updated.

[0054] Specifically, the loss is calculated using the following formula: ; in, represents the loss of the masked feature frame in the sample, represents the loss of the unmasked feature frame in the sample, represents the weight, .

[0055] In implementation, due to the random masking, the loss is calculated separately for masked feature frames and unmasked feature frames.

[0056] Specifically, the loss of the masked feature frame is calculated using the following formula: ; in, represents the clustering type loss of the i-th masked feature frame, represents the distance loss of the i-th masked feature frame, Indicates the number of masked feature frames in the sample.

[0057] Specifically, the clustering type loss of the i-th masked feature frame is calculated using the following formula: ; in, represents the hidden representation of the i-th masked feature frame, represents the label of the cluster type of the i-th masked feature frame, C represents the number of cluster types, represents the scaling parameter, A represents the projection matrix, Indicates the label of the j-th cluster type.

[0058] The cluster type label of the i-th masked feature frame is the cluster type label of the corresponding signal frame.

[0059] For the i-th masked feature frame, the distance to the feature frames of the same type should be close, while the distances to feature frames of different types are far. Therefore, the following formula is used to calculate the distance loss of the i-th masked feature frame: ; in, represents the hidden representation of the i-th masked feature frame, represents the hidden representation of the p-th feature frame, represents the hidden representation of the qth feature frame; the cluster type label of the pth feature frame is the same as the cluster type label of the i-th masked feature frame, and the cluster type label of the qth feature frame is different from the cluster type label of the i-th masked feature frame. Represents the margin parameters.

[0060] Used to represent the minimum distance difference between feature frames of different clustering types.

[0061] By considering label loss and clustering loss, the classification accuracy of the model is improved.

[0062] The loss of the unmasked feature frame is calculated in the same way , based on the loss, the parameters of the pre-trained model are updated in reverse order, and when the training stop condition is reached, the training is stopped to obtain a trained pre-trained model. The training stop condition may be, for example, reaching a preset loss accuracy or the number of training times.

[0063] After obtaining the trained pre-trained model, the convolutional coding module, the BERT coding module and the classification module constitute a bowel sound classification model. The parameters of the convolutional coding module are fixed, and the bowel sound classification model is fine-tuned based on the training sample set to obtain a trained bowel sound classification model, wherein the projection matrix of the BERT coding module is removed and replaced with a randomly initialized softmax layer. The classification module predicts the type of bowel sound signal based on the occurrence location of the bowel sound and the hidden representation of each feature frame.

[0064] During fine-tuning, the classification loss can be calculated through the cross entropy loss function, and the parameters of the BERT encoding module and the classification module can be updated. When the training stop condition is reached, the training is stopped to obtain a trained bowel sound classification model.

[0065] The bowel sound classification module inputs the location of bowel sounds and bowel sound signals into the trained bowel sound classification model to quickly and accurately obtain the individual's bowel sound type.

[0066] Those skilled in the art will appreciate that all or part of the processes of the above-mentioned embodiments can be implemented by instructing related hardware through a computer program, and the program can be stored in a computer-readable storage medium, wherein the computer-readable storage medium is a disk, an optical disk, a read-only storage memory, or a random access memory, etc.

[0067] The above description is only a preferred specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. Any changes or substitutions that can be easily conceived by any technician familiar with the technical field within the technical scope disclosed by the present invention should be covered within the protection scope of the present invention.

Claims

1. A bowel sound detection system, characterized in that: The system comprises: The signal acquisition device comprises a laser, a distributed optical fiber sensor and a collector; the laser is used to periodically emit laser light to the distributed optical fiber sensor; the collector is used to collect the signal returned by the optical fiber sensor and send it to the bowel sound solution module; The bowel sound calculation module is used to calculate the collected signal based on optical frequency domain reflection to determine whether bowel sounds occur. If so, the location of the bowel sounds is determined and the bowel sound signals are extracted; The bowel sound classification module is used to obtain the individual bowel sound type based on the location of the bowel sound and the bowel sound signal based on the trained bowel sound classification model.

2. The bowel sound detection system according to claim 1, characterized in that: The distributed optical fiber sensor comprises a test optical fiber, a reference optical fiber, a first coupler and a circulator; the first coupler is connected to a laser and is used to divide the light source into two optical paths, one of which is injected into the reference optical fiber and the other is injected into the test optical fiber through the circulator; The collector includes a second coupler and a photoelectric detector, wherein the second coupler is connected to the reference optical fiber and the circulator, and is used to couple the signals returned by the reference optical fiber and the test optical fiber and send them to the photoelectric detector; The photoelectric detector converts the optical signal into an electrical signal and sends it to the bowel sound calculation module.

3. The bowel sound detection system according to claim 2, characterized in that: The signal acquisition device also includes a cylindrical shell; the inner wall of the cylindrical shell is provided with an annular groove; the test optical fiber of the distributed optical fiber sensor is wound on the inner wall of the cylindrical shell, and the test optical fiber of the distributed optical fiber sensor is embedded in the groove of the inner wall of the cylindrical shell.

4. The bowel sound detection system according to claim 2, characterized in that: The bowel sound calculation module calculates the collected signal in the following manner to determine whether bowel sounds occur, and if so, to determine the location of the bowel sounds: Acquire a reference signal; the reference signal has the same length as the acquisition signal; Whether bowel sounds occur is determined based on the similarity between the reference signal and the collected signal, and if so, the location of the bowel sounds is located based on the similarity.

5. The bowel sound detection system according to claim 4, characterized in that: The following method is used to determine whether bowel sounds occur based on the similarity between the reference signal and the collected signal: The reference signal and the acquisition signal are segmented respectively according to the emission period of the laser; The reference signal and each segmented signal of the acquired signal are converted into the range domain by fast Fourier transform; For each distance domain segment of the collected signal, the distance domain segment of the collected signal and the corresponding distance domain segment of the reference signal are respectively divided into a plurality of subdivided segments by using a sliding window; For each subdivision of the distance domain segmentation, the subdivision and the subdivision of the reference signal corresponding to the subdivision are respectively converted into time domain signals by using inverse Fourier transform; the cross-correlation diagram of the two groups of time domain signals is calculated to obtain the similarity of the subdivision; If there is a subdivision whose similarity is less than the first threshold in the distance segment, it is determined that bowel sounds have occurred.

6. The bowel sound detection system according to claim 1, characterized in that: The bowel sound classification model includes: A convolutional coding module is used to extract features from the bowel sound signal to obtain a feature frame sequence; the feature frame sequence is randomly masked and input into a BERT coding module; BERT encoding module, which is used to output the hidden representation of each feature frame based on the sequence of feature frames after random masking; A classification module is used to predict the type of bowel sounds based on the location of the bowel sounds and the hidden representation of each feature frame.

7. The bowel sound detection system according to claim 3, characterized in that: The trained bowel sound classification model is obtained in the following way: Acquire multiple bowel sound signals to construct a training sample set; A convolutional coding module and a BERT coding module are combined into a pre-training model; a clustering module is added to the pre-training model; the clustering module is used to cluster the signal frames of the bowel sound signal to obtain a label of each cluster type; the BERT coding module also outputs a prediction result of the cluster type of each feature frame; The pre-trained model is trained based on the label of the clustering type of each feature frame of the sample and the predicted clustering type calculation loss to obtain a trained pre-trained model; The parameters of the convolutional coding module are fixed, and the bowel sound classification model is fine-tuned based on the training sample set to obtain a trained bowel sound classification model.

8. The bowel sound detection system according to claim 4, characterized in that: The loss is calculated using the following formula: ； in, represents the loss of the masked feature frame in the sample, represents the loss of the unmasked feature frame in the sample, Represents weight.

9. The bowel sound detection system according to claim 4, characterized in that: The loss of the masked feature frame is calculated using the following formula: ； in, represents the clustering type loss of the i-th masked feature frame, represents the distance loss of the i-th masked feature frame, Indicates the number of masked feature frames in the sample.

10. The bowel sound detection system according to claim 9, characterized in that: The distance loss of the i-th masked feature frame is calculated using the following formula: ； in, represents the hidden representation of the i-th masked feature frame, represents the hidden representation of the p-th feature frame, represents the hidden representation of the qth feature frame; the cluster type label of the pth feature frame is the same as the cluster type label of the i-th masked feature frame, and the cluster type label of the qth feature frame is different from the cluster type label of the i-th masked feature frame. Represents the margin parameters.

Images