JP6672478B2 - Body sound analysis method, program, storage medium, and body sound analysis device - Google Patents

Description

  The present invention relates to a technical field of a body sound analysis method, a program, a storage medium, and a body sound analysis device for analyzing a body sound such as a breathing sound.

  2. Description of the Related Art There has been known an apparatus that detects an abnormal sound (that is, a sound different from a normal respiratory sound) contained in a respiratory sound of a living body detected by an electronic stethoscope or the like. For example, Patent Literature 1 describes a technique of detecting a plurality of abnormal sounds (sub-noises) included in breathing sounds by decomposing the sounds into sound types.

International Publication No. WO 2016/002004

  When analyzing the abnormal sound included in the body sound, the acquired body sound information and the previously stored abnormal sound information (specifically, the body sound when the abnormal sound is actually generated) ) Can be determined to be an abnormal sound. However, since the body sound information fluctuates according to individual differences, measurement environments, and the like, it is difficult to accurately determine whether or not an abnormal sound is occurring simply by comparing the information. For this reason, unless an appropriate judgment criterion is set, there arises a technical problem that the abnormal sound cannot be detected even though the abnormal sound is generated, or is erroneously detected when the abnormal sound is not generated.

  The problems to be solved by the present invention include, for example, those described above. An object of the present invention is to provide a body sound analysis method, a program, a storage medium, and a body sound analysis device capable of suitably analyzing abnormal sounds included in body sounds.

  A body sound analysis method for solving the above-mentioned problem is a body sound analysis method used in a body sound analysis device for analyzing a body sound, wherein a first acquisition step of acquiring first information regarding a body sound; A second acquisition step of acquiring second information indicating a timing at which an abnormal sound is occurring in the first information; a learning step of learning a correspondence relationship between the first information and the second information; Determining the abnormal sound included in the input body sound information based on the learning result by the above.

  A program for solving the above-described problem causes the body sound analysis device to execute the above-described body sound analysis method.

  A storage medium for solving the above-mentioned problem stores the above-mentioned program.

  A body sound analysis device for solving the above-mentioned problem includes a first acquisition unit that acquires body sound information related to body sound, and a determination unit that determines an abnormal sound included in the body sound based on a learning result. The learning result is a correspondence relationship between the first information and the second information based on first information on the body sound and second information indicating a timing at which an abnormal sound is generated in the body sound. Is the learning result of learning.

FIG. 3 is a block diagram illustrating a configuration of a frame determination learning device according to the embodiment. 5 is a flowchart illustrating a flow of a frame determination learning operation according to the embodiment. It is a conceptual diagram which shows the frame division process of a teacher audio | voice signal. It is a flowchart which shows the calculation process of a 1st local feature-value. It is a flowchart which shows the calculation process of a 2nd local feature-value. FIG. 4 is a diagram illustrating a local feature vector obtained from a waveform and a spectrum. It is a figure showing teacher frame information matched with a local feature. FIG. 4 is a block diagram illustrating a configuration of a threshold value determining unit according to the embodiment. It is a conceptual diagram which shows an example of the processing content of ROC analysis. 6 is a flowchart illustrating a flow of an optimal threshold value determining operation according to the embodiment.

<1>
The body sound analysis method according to the present embodiment is a body sound analysis method used in a body sound analysis device that analyzes body sound, and includes a first acquisition step of acquiring first information related to body sound, A second acquisition step of acquiring second information indicating a timing at which an abnormal sound occurs in the information, a learning step of learning a correspondence relationship between the first information and the second information, and a learning by the learning step Discriminating the abnormal sound included in the input body sound information based on the result.

  According to the body sound analysis method according to the present embodiment, the first information on the body sound (for example, respiratory sound) is obtained, and the timing at which the abnormal sound (for example, auxiliary noise) in the first information is generated is determined. The indicated second information is obtained. The first information is acquired as information indicating a temporal change of the body sound (for example, a time axis waveform indicating the body sound). On the other hand, it is desired that the second information is information that accurately indicates the timing at which the abnormal sound in the first information is occurring. For this reason, the second information is preferably information prepared in advance using the first information.

  When the first information and the second information are obtained, the correspondence between the first information and the second information is learned. Specifically, a parameter for deriving the second information from the first information is learned. There may be more than one type of this parameter. The learning step is preferably performed a plurality of times using a plurality of pieces of first information and a plurality of pieces of second information in order to make the learning result more accurate.

  After the learning step, an abnormal sound included in the input body sound information is determined based on the learning result. The “input body sound information” is information on a body sound to be analyzed by the body sound analysis method according to the present embodiment, and is input separately from the above-described first information and second information. is there. In the present embodiment, in particular, since the correspondence between the first information and the second information is learned in advance, it is possible to accurately determine the timing at which the abnormal sound is generated from the input body sound information. Therefore, it is possible to appropriately determine the abnormal sound included in the body sound information.

<2>
In one aspect of the body sound analysis method according to the present embodiment, the method further includes a first generation step of generating feature amount information indicating a feature amount in the first information based on the first information, and the learning step includes: The correspondence between the feature amount information and the second information is learned instead of the correspondence between the first information and the second information.

  According to this aspect, when the first information is obtained, feature amount information indicating a feature amount in the first information is generated. The “feature amount” is a value indicating a magnitude (degree) of a feature that can be used to determine an abnormal sound included in the body sound.

  In this aspect, in particular, the correspondence between the feature amount information and the second information is learned instead of the correspondence between the first information and the second information. Therefore, a more suitable learning result can be obtained in order to determine an abnormal sound included in the input body sound information.

<3>
In one aspect of the body sound analysis method according to the present embodiment, the method further includes a dividing step of dividing the first information and the second information into predetermined frame units, and the learning step learns in the predetermined frame units. .

  According to this aspect, while learning is performed, the first information and the second information are divided into predetermined frame units. The predetermined frame unit is set as a period in which an appropriate learning result can be more easily obtained. For this reason, by performing learning in a predetermined frame unit, it is possible to more appropriately obtain a learning result.

<4>
In one aspect of the body sound analysis method according to the present embodiment, a third obtaining step of obtaining third information indicating whether or not the abnormal sound has occurred in the first information, and the first information and the learning step include: A calculating step of calculating, based on a learning result, fourth information indicating a ratio of a period during which the abnormal sound is generated to a period during which the first information is obtained, and calculating the third information and the fourth information. And determining a threshold value for determining whether or not the input body sound information includes an abnormal sound based on the input sound information.

  According to this aspect, the third information indicating whether or not the abnormal sound has occurred in the first information is obtained. It is desired that the third information is information that accurately indicates whether or not an abnormal sound has occurred in the first information. For this reason, it is preferable that the third information is information prepared in advance using the first information.

  Further, in the present aspect, based on the first information and the learning result, fourth information indicating a ratio of a period during which the abnormal sound is generated to a period during which the first information is obtained is calculated. Specifically, the fourth information is calculated by analyzing the first information using the learning result.

  When the third information is obtained and the fourth information is calculated, a threshold for determining whether or not the input body sound information includes an abnormal sound is determined based on the third information and the fourth information. You. This threshold value is a threshold value for determining whether or not an abnormal sound is actually included when determining an abnormal sound included in the input body sound information. This is a threshold value that is compared with a value indicating a ratio of a period during which abnormal sound is generated to a period during which information is acquired.

  By using this threshold value, it is possible to determine that an abnormal sound has occurred, for example, when the ratio of the period during which the abnormal sound related to the input body sound information is occurring is equal to or greater than the determined threshold value. On the other hand, when the ratio of the period during which the abnormal sound related to the input body sound information is generated is less than the determined threshold, it can be determined that the abnormal sound is not generated.

  Particularly in this aspect, since the threshold is determined based on the fourth information calculated using the learning result, it is possible to more accurately determine whether or not an abnormal sound has occurred.

<5>
The program according to the present embodiment causes the body sound analyzer to execute the above-described body sound analysis method.

  According to the program according to the present embodiment, since the above-described body sound analysis method according to the present embodiment can be executed, it is possible to appropriately determine an abnormal sound included in body sound information.

<6>
The storage medium according to the present embodiment stores the above-described program.

  According to the storage medium according to the present embodiment, the program according to the above-described embodiment can be executed, so that abnormal sounds included in body sound information can be appropriately determined.

<7>
The body sound analysis device according to the present embodiment includes a first acquisition unit that acquires body sound information related to body sound, and a determination unit that determines an abnormal sound included in the body sound based on a learning result, The learning result learns a correspondence relationship between the first information and the second information based on first information related to the body sound and second information indicating a timing at which an abnormal sound occurs in the body sound. It is the learning result that was done.

  According to the body sound analysis apparatus according to the present embodiment, similarly to the above-described body sound analysis method, it is possible to appropriately determine the abnormal sound included in the body sound information based on the learning result.

  Note that the body sound analysis device according to the present embodiment can also adopt various aspects similar to the various aspects in the above-described body sound analysis method according to the embodiment.

  The operation and other gains of the body sound analysis method, the program, the storage medium, and the body sound analysis device according to the present embodiment will be described in more detail in the following examples.

  Hereinafter, embodiments of a body sound analysis method, a program, a storage medium, and a body sound analysis device will be described in detail with reference to the drawings. Hereinafter, a body sound analysis method for analyzing a breathing sound will be described as an example.

<Teacher data>
First, teacher data used in the body sound analysis method according to the present embodiment will be described.

  The teacher data is data in which three pieces of information of a teacher voice signal, teacher frame information, and entire teacher information are set as one set, and a plurality of sets are prepared in advance.

  The teacher voice signal is a signal (for example, a time axis waveform) indicating a temporal change of the breathing sound. The teacher frame information is information indicating the generation timing of the abnormal sound in the teacher voice signal for each sound type. The entire teacher information is information indicating whether or not an abnormal sound has occurred in the teacher voice signal for each sound type.

  The teacher data is used for a learning operation described later, and the greater the number, the higher the learning effect (in other words, the accuracy of the body sound analysis).

<Frame judgment learning>
Next, frame determination learning of the body sound analysis method according to the present embodiment will be described with reference to FIGS. Note that the frame determination learning is a learning operation for improving the determination accuracy of the frame determination process for determining the occurrence of abnormal sound on a frame-by-frame basis.

<Structure of learning device>
First, the configuration of a frame determination learning device used for frame determination learning will be described with reference to FIG. FIG. 1 is a block diagram illustrating a configuration of the frame determination learning device according to the present embodiment.

  As shown in FIG. 1, the frame determination learning device according to the present embodiment includes a teacher voice signal input unit 110, a teacher frame information input unit 120, a processing unit 200, and a learning result output unit 300. ing.

  The teacher voice signal input unit 110 is configured to acquire a teacher voice signal included in teacher data and output it to the processing unit 200.

  The teacher frame information input unit 120 is configured to acquire teacher frame information included in teacher data and output the acquired teacher frame information to the processing unit 200.

  The processing unit 200 includes a plurality of arithmetic circuits, memories, and the like. The processing unit 200 includes a frame dividing unit 210, a first local feature calculating unit 220, a frequency analyzing unit 230, a second local feature calculating unit 240, and a learning unit 250.

  The frame division unit 210 is configured to be capable of executing a division process of dividing the teacher voice signal input from the teacher voice signal input unit 110 into a plurality of frames. The respiratory sound signal divided by the frame division unit 210 is output to the first local feature value calculation unit 220 and the frequency analysis unit 230.

  The first local feature calculating section 220 is configured to be able to calculate the first local feature based on the waveform of the teacher voice signal. The processing executed by the first local feature calculating unit 220 will be described later in detail. The first local feature value calculated by the first local feature value calculation unit 220 is configured to be output to the learning unit 250.

  The frequency analysis unit 230 is configured to be able to execute a time-frequency analysis process (for example, an FFT process) on the teacher voice signal input from the teacher voice signal input unit 110. The analysis result of the time-frequency analyzer 230 is output to the second local feature calculator 240.

  The second local feature calculation unit 240 is configured to be able to calculate the second local feature based on the analysis result of the frequency analysis unit 230. The processing executed by the second local feature value calculation unit 240 will be described later in detail. The second local feature calculated by the second local feature calculator 240 is output to the learning unit 250.

  The learning unit 250 learns the correspondence between the local feature amounts calculated by the first local feature amount calculation unit 220 and the second local feature amount calculation unit 240 and the teacher frame information input from the teacher frame information input unit 120. It is configured to be possible. The processing executed by the learning unit 250 will be described later in detail. The learning result of the learning unit 250 is configured to be output to the learning result output unit 300.

  The learning result output unit 300 is configured to be able to output the learning result of the learning unit 250 in a form that can be used for analyzing the body sound. For example, the learning result output unit 300 is configured to be able to output the learning result of the learning unit 250 to a memory or the like of the body sound analyzer.

<Operation description>
Next, a flow of a frame determination learning operation performed by the above-described frame determination learning device will be described with reference to FIG. FIG. 2 is a flowchart illustrating the flow of the frame determination learning operation according to the present embodiment.

  As shown in FIG. 2, at the time of the frame learning operation according to the present embodiment, first, the learning unit 250 is initialized (step S100). Subsequently, a teacher voice signal is acquired by the teacher voice signal input unit 110 (step S101). The teacher voice signal input unit 110 outputs the acquired teacher voice signal to the processing unit 200. The teacher voice signal is a specific example of “first information”.

  Subsequently, the respiratory sound is divided into a plurality of frames by the frame dividing unit 210 (step S102). Hereinafter, the frame division of the respiratory sound signal will be specifically described with reference to FIG. FIG. 3 is a conceptual diagram showing a frame division process of the teacher voice signal.

  As shown in FIG. 3, the teacher voice signal is divided into a plurality of frames at predetermined intervals. This frame is set as a processing unit for suitably executing a later-described local feature amount calculation process, and a period per frame is, for example, 12 msec.

  Returning to FIG. 2, the frame-divided teacher voice signal is input to the first local feature calculator 220, and the first local feature is calculated (step S103). Further, the teacher voice signal that has been divided into frames is frequency-analyzed by the frequency analysis unit 230, and is input to the second local feature value calculation unit 240. The second local feature calculating unit 240 calculates a second local feature based on the frequency-analyzed teacher voice signal (for example, a spectrum indicating frequency characteristics).

  In the following, the calculation process of the first local feature by the first local feature calculator 220 and the calculation of the second local feature by the second local feature calculator 240 will be described with reference to FIGS. This will be described in detail. FIG. 4 is a flowchart illustrating a process of calculating the first local feature value. FIG. 5 is a flowchart illustrating a calculation process of the second local feature value. FIG. 6 is a diagram showing a local feature vector obtained from a waveform and a spectrum.

  As shown in FIG. 4, when calculating the first local feature value, first, the waveform of the teacher voice signal is obtained (step S201), and a pre-filter process is performed (step S202). The pre-filter process is, for example, a process using a high-pass filter, and can remove an extra component included in the teacher voice signal.

  Subsequently, a local variance value is calculated using the teacher sound signal that has been subjected to the pre-filter processing (step S203). The local variance value is calculated, for example, as a first variance value indicating a variation in the teacher voice signal in the first period w1 and a second variance value indicating a variation in the teacher voice signal in the second period w2 including the first period w1. . The local variance value calculated in this manner functions as a local feature value for determining intermittent rales (for example, water bubbles) among abnormal sounds.

  When the local variance value is calculated, the maximum value of the local variance value in each frame of the teacher voice signal is calculated (step S204), and output as the first local feature value (step S205).

  As shown in FIG. 5, when calculating the second local feature, first, a spectrum obtained by frequency analysis is obtained (step S301), and a CMN (Cepstral Mean Normalization) process is executed (step S302). In the CMN process, it is possible to remove the constantly convoluted characteristics such as the sensor and the environment from the teacher voice signal.

  A lifter process (step S303) for extracting an envelope component and a lifter process (step S304) for extracting a fine component are further performed on the teacher voice signal subjected to the CMN process. The liftering process is a process of cutting a predetermined cffrency component from the cepstrum.

  According to the above-described CMN processing and liftering processing, it is possible to make a state in which a continuous rattle (for example, a snoring sound, a whistling sound, a crouching sound, etc.) buried in other body sounds can be easily distinguished. Note that the CMN processing and the lifting processing are existing techniques, and therefore, a more detailed description thereof will be omitted.

  With respect to the teacher voice signal on which the lifter processing for extracting the fine component has been performed, the enhancement processing using the KL information amount is executed, and the feature amount is calculated. The KL information amount is a parameter calculated using the observed value P and a reference value Q (for example, a theoretical value, a model value, a predicted value, etc.), and an observed value P having a characteristic with respect to the reference value Q appears. And the KL information amount is calculated as a large value. According to the process using the KL information amount, the tone component (that is, the component for determining the continuous ra tone) included in the teacher voice signal is emphasized and clarified.

  On the other hand, the calculation of the HAAR-LIKE feature is also performed on the spectrum obtained by the frequency analysis (step S306). The HAAR-LIKE feature is a technique mainly used in the field of image processing. Here, the HAAR-LIKE feature is calculated from a spectrum by a similar method by associating an amplitude value for each frequency with a pixel value in image processing. Note that the calculation of the HAAR-LIKE feature is an existing technique, and a detailed description thereof will be omitted.

  As described above, when a plurality of feature values are calculated for each frequency by performing various processes on the teacher voice signal, an average value is calculated for each frequency band (step S307) and output as a second local feature value (step S307). Step S307).

  As shown in FIG. 6, a plurality of types of local feature amounts (that is, a first local feature amount and a second local feature amount) are obtained from the waveform and the spectrum of the respiratory sound signal by the above-described processing. These are output as local feature vectors for each frame.

  Returning to FIG. 2 again, when the local feature amount is calculated, subsequently, the teacher frame information input unit 120 acquires the teacher frame information (step S106). The teacher frame information is a specific example of “second information”. The acquired teacher frame information is output to the learning unit 250 together with the local feature amount.

  Hereinafter, the learning operation in the learning unit 250 will be described with reference to FIG. FIG. 7 is a diagram illustrating teacher frame information associated with local feature amounts.

  As shown in FIG. 7, in the learning unit 250, the local feature amount and the teacher frame information are set as paired data for each frame (step S107). Thus, the local feature amount and the timing of occurrence of the abnormal sound are associated with each other.

  More specifically, the local feature value vector associated with the teacher frame information indicating the timing at which the abnormal sound is generated is learned as the local feature value when the abnormal sound is generated. On the other hand, the local feature value vector associated with the teacher frame information indicating the timing at which no abnormal sound is generated is learned as the local feature value when no abnormal sound is generated.

  Returning to FIG. 2 again, since the association between the local feature amount and the teacher frame information is performed for each frame, when the setting is completed, it is determined whether or not the setting has been completed for all the frames (step S108). If the setting has not been completed for all the frames (step S108: NO), the processing from step S103 and S104 is repeated again for the unset frames.

  On the other hand, when the setting has been completed for all the frames (step S108: YES), it is determined whether or not the processing has been completed for all the plurality of teacher data (step S109). If it is determined that the processing has not been completed for all the teacher data (step S109: NO), the processing from step S101 is executed again. By repeating the processing in this way, all the frames of the teacher data are associated with each other.

  Then, the learning process is actually performed by the learning unit 250 (step S110). The learning process is performed using a machine learning algorithm such as AdaBoost. In the learning process, an existing method other than the above-described AdaBoost can be appropriately used, and thus a detailed description thereof is omitted.

  Since the above-described learning processing is performed for each abnormal sound type, it is determined whether or not the learning processing has been completed for all sound types after the processing is completed (step S111), and the learning processing has been completed for all sound types. If not (step S111: NO), the learning process of step S110 is executed again for uncompleted tone types.

  On the other hand, when the learning process has been completed for all the sound types (step S111: YES), the learning result is output to the body sound analyzer (step S112).

<Effects of frame judgment learning>
The learning result of the above-described frame determination learning is used for analysis of respiratory sounds by the body sound analyzer. At the time of analyzing the breathing sound, the same processing as that for the teacher voice signal in the above-described frame determination learning operation is performed on the breathing sound signal to be analyzed. Specifically, the input respiratory sound signal is divided into frames, and a local feature value is calculated for each frame.

  In the present embodiment, since the relationship between the local feature value and the occurrence timing of the abnormal sound is learned by the above-described learning operation, the generation of the abnormal sound is preferably performed using the local feature value calculated from the respiratory sound. The timing (in other words, the temporal position of the frame where the abnormal sound occurs) can be detected. Therefore, the abnormal sound can be determined very accurately as compared with the case where the learning operation is not performed in advance.

<Determination of optimal threshold>
Next, the optimal threshold value determining operation according to the present embodiment will be described with reference to FIGS. 8 and 10. Note that the optimal threshold value determining operation is defined as the ratio of the period during which it is determined that an abnormal sound is occurring. This is an operation for determining a threshold value used for determining whether or not an actual abnormal sound is actually generated as an optimum value based on the threshold value.

<Configuration of threshold value determination unit>
First, the configuration of the threshold value determining unit that performs the optimal threshold value determining operation will be described with reference to FIG. FIG. 8 is a block diagram illustrating a configuration of the threshold value determination unit according to the present embodiment.

  As illustrated in FIG. 8, the threshold determination unit according to the present embodiment includes a frame determination result input unit 410, an entire teacher information input unit 420, a determination unit 500, and a threshold output unit 600. .

  The frame determination result input unit 410 acquires the result of the frame determination process of the teacher voice signal using the learning result (that is, the process of determining whether or not an abnormal sound is generated for each frame), and determines the determination unit 500. It is configured to be able to output to.

  The teacher entire information input unit 120 is configured to acquire the teacher entire information included in the teacher data and output the acquired teacher information to the determination unit 500.

  The determination unit 500 includes a global feature amount calculation unit 510, an ROC analysis (Receiver Operating Characteristic analysis) unit 520, and an optimum threshold value calculation unit 530.

  The global feature amount calculation unit 510 is configured to be able to calculate the ratio of the occurrence time during which the abnormal sound is generated to the period during which the teacher voice signal is input, based on the determination result of the frame determination process. Information indicating the ratio of the abnormal sound occurrence time calculated by the global feature amount calculation unit 510 is output to the ROC analysis unit 520.

  The ROC analysis unit 520 determines, based on the relationship between the information indicating the ratio of the abnormal sound occurrence time and the entire teacher information, a certain threshold value for the ratio of the abnormal sound occurrence time and a determination using the threshold value. An ROC analysis for acquiring a relationship with performance as an ROC curve is configured to be executable. Since the ROC analysis is an existing technique, a detailed description is omitted here. The analysis result of the ROC analysis unit 520 is output to the optimum threshold value calculation unit 530.

  The optimum threshold value calculation unit 530 uses the analysis result of the ROC analysis unit 520 to determine, from the ratio of the occurrence time of the abnormal sound, whether or not the abnormal sound is actually occurring. A threshold (a threshold that gives a point closest to the reference point (0, 1) in the ROC curve) can be calculated (see FIG. 9). The threshold calculated by the optimum threshold calculator 530 is output to the threshold output unit 600.

  The threshold value output unit 600 is configured to be able to output the threshold value calculated by the optimum threshold value calculation unit 530 in such a manner that it can be used for analysis of body sounds. For example, the threshold output unit 300 is configured to be able to output the threshold calculated by the optimum threshold calculator 530 to a memory or the like of the body sound analyzer.

<Operation description>
Next, the flow of the optimal threshold value determining operation performed by the above-described threshold value determining unit will be described with reference to FIG. FIG. 10 is a flowchart illustrating the flow of the optimum threshold value determining operation according to the present embodiment.

  As shown in FIG. 10, at the time of the optimum threshold value determining operation according to the present embodiment, first, the frame determination result of the teacher voice signal is acquired by the frame determination result input unit 410 (step S201). Subsequently, it is determined whether or not frame determination results for all frames have been obtained (step S202). If the frame determination results for all the frames have not been acquired (step S202: NO), the process of step S201 is executed again for the frames that have not been acquired.

  On the other hand, if the frame determination results of all the frames have been obtained (step S202: YES), the obtained frame determination results are output to the global feature amount calculation unit 510, and the frames of the entire section of the teacher voice signal are output. The ratio of the number of frames determined to have an abnormal sound to the number (that is, global feature amount) is calculated (step S203). The global feature is a specific example of “fourth information”.

  Subsequently, the entire teacher information included in the teacher data is acquired by the entire teacher information input unit 420 (step S204). The teacher entire information is a specific example of “third information”.

  The overall teacher information is output to the ROC analysis unit 520 together with the global feature amount, and the global feature amount and the overall teacher information are set as a pair of data (step S205). That is, the ratio of the number of frames determined to have an abnormal sound is associated with information indicating whether or not an abnormal sound has occurred. Thereafter, it is determined whether or not the setting has been completed for all the teacher data (step S206). If the setting of all the teacher data has not been completed (step S206: NO), the processing from step S201 is executed again on the unset teacher data.

  On the other hand, if the setting of all the teacher data has been completed (step S206: YES), the ROC analysis is performed by the ROC analysis unit 520, and the relationship between the threshold and the discrimination performance is obtained as an ROC curve (step S207). . When the ROC analysis ends, the optimum threshold value is calculated by the optimum threshold value calculation unit 530 according to the ROC analysis result (step S208).

  Since the above-described ROC analysis and threshold value calculation processing are performed for each sound type of the abnormal sound, it is determined whether or not the processing has been completed for all sound types after the end (step S209). If the processing has not been completed for all tone types (step S209: NO), the processing of steps S207 and S208 is executed for uncompleted tone types. If the processing has been completed for all tone types (step S209: YES), a threshold is output (step S210), and the series of processing ends.

<Effect of determining the optimal threshold>
The threshold value determined as described above is used for analysis of respiratory sounds by the body sound analyzer. For example, when analyzing a breathing sound, the same processing as that for the teacher voice signal in the above-described frame determination learning operation is performed on the breathing sound signal to be analyzed. Specifically, the ratio of the period during which the abnormal sound is generated to the period during which the breathing sound is acquired (that is, the global feature amount) is calculated from the frame determination result of the input breathing sound signal. As a result, for example, when the ratio of the period during which the abnormal sound is generated is large, it can be determined that the abnormal sound is actually generated. On the other hand, when the ratio of the period during which the abnormal sound is generated is small, it can be determined that the abnormal sound is not actually generated although the abnormal sound is detected in the frame unit.

  The determination regarding the presence or absence of such an abnormal sound is realized by comparison with the threshold determined by the above-described optimal threshold determination operation. Specifically, when the ratio of the period during which the abnormal sound occurs is greater than the threshold, it can be determined that the abnormal sound has occurred, and when it is smaller than the threshold, it can be determined that the abnormal sound has not occurred. Here, in the present embodiment, in particular, in the above-described optimal threshold value determining operation, the global feature amount and the entire teacher data (that is, information indicating whether or not an abnormal sound is generated) are associated with each other. Is calculated. Therefore, the presence or absence of occurrence of an abnormal sound can be determined extremely accurately based on the global feature amount calculated from the breathing sound.

  The present invention is not limited to the above-described embodiment, and can be appropriately modified within a scope not contrary to the gist or idea of the invention which can be read from the claims and the entire specification. The method, the program, the storage medium, and the body sound analyzer are also included in the technical scope of the present invention.

Reference Signs List 110 teacher voice signal input unit 120 teacher frame information input unit 200 processing unit 210 frame division unit 220 first local feature calculation unit 230 frequency analysis unit 240 second local feature calculation unit 250 learning unit 300 learning result output unit 410 frame determination Result input unit 420 Teacher whole information input unit 500 Determination unit 510 Global feature amount calculation unit 520 ROC analysis unit 530 Optimal threshold calculation unit 600 Threshold output unit

Claims (6)

A body sound analysis method used in a body sound analyzer for analyzing body sound,
A first acquisition step of acquiring first information that is information indicating a temporal change of a body sound;
A second acquisition step of acquiring second information indicating a timing at which an abnormal sound is occurring in the first information;
A learning step of learning a correspondence between the first information and the second information;
A determining step of determining an abnormal sound included in the input body sound information based on a learning result of the learning step;
A third obtaining step of obtaining third information indicating whether or not the abnormal sound has occurred in the first information;
A calculating step of calculating, based on the first information and the learning result obtained by the learning step, fourth information indicating a ratio of a period during which the abnormal sound is generated to a period during which the first information is obtained;
Based on the third information and the fourth information, a determining step of determining a threshold for determining whether the input body sound information includes an abnormal sound,
A body sound analysis method, comprising: A first generation step of generating feature amount information indicating a feature amount in the first information based on the first information,
The living body according to claim 1, wherein the learning step learns a correspondence between the feature amount information and the second information instead of a correspondence between the first information and the second information. Sound analysis method. A dividing step of dividing the first information and the second information into predetermined frame units;
The body sound analysis method according to claim 1, wherein the learning step performs learning in units of the predetermined frame.   A program for causing the body sound analyzer to execute the body sound analysis method according to claim 1.   A storage medium storing the program according to claim 4. A first acquisition unit that acquires body sound information related to body sounds,
A determination unit configured to determine an abnormal sound included in the body sound based on a learning result;
With
The learning result is based on first information that is information indicating a temporal change of the body sound, and second information indicating a timing at which an abnormal sound in the body sound is generated. It is a learning result of learning a correspondence relationship with the second information,
A third obtaining unit that obtains third information indicating whether or not the abnormal sound has occurred in the first information;
A calculating unit configured to calculate, based on the first information and the learning result, fourth information indicating a ratio of a period during which the abnormal sound is generated to a period during which the first information is acquired;
A determination unit based on the third information and the fourth information, determines a threshold value for determining whether or not included before Kisei body sounds abnormal sound information,
A biological sound analyzer, further comprising:

Images