JP2014090722A - Breath sound analyzer, breath sound analyzing method and breath sound analyzing program - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a breath sound analyzer, capable of calculating an evaluation value for determining whether or not an abnormal sound is contained in the breath sound and an evaluation value for determining the kind of the abnormal sound in a simple method.SOLUTION: A peak section identification part 120 divides an input signal based on a breath sound into a plurality of frequency bands, and generates a peak section signal showing a relatively high level section and a non-peak section signal showing the section other than the peak section in signals of the respective frequency bands. Further a peak section composite signal obtained by synthesizing the plurality of peak section signals and a non-peak section composite signal obtained by synthesizing the plurality of non-peak section signals are generated. A feature quantity generating part 130 generates a first feature quantity from the peak section composite signal, and generates a second feature quantity from the non-peak section composite signal. An evaluation value calculating part 140 calculates a first evaluation value for determining whether or not an abnormal sound is contained in the breath sound and a second evaluation value for determining the kind of the abnormal sound using the first feature quantity and the second feature quantity.

Description

  The present invention relates to a technique for analyzing respiratory sounds and determining whether or not abnormal sounds are included.

Various information is contained in the breath sound. The doctor diagnoses the presence or absence of abnormality and the type of abnormality by listening to breathing sounds using a stethoscope. On the other hand, in recent years, expansion of home medical care has been demanded, and a device capable of analyzing breathing sounds is required even if the doctor has no specialized knowledge. Therefore, if it is possible to analyze respiratory sounds in a computer or other device and determine the presence or absence and type of abnormality, it is considered that this will lead to expansion of home medical care.
Patent Document 1 discloses a respiratory condition analyzer that identifies an input signal as a breathing sound interval and a non-breathing sound interval, and identifies whether the breathing sound interval is a normal breathing sound or an abnormal breathing sound. ing.

JP 2012-120688 A

In diagnosis of respiratory sounds by a doctor, generally, the presence or absence of abnormality is determined from the overall tendency of respiratory sounds. Rather than judging that an abnormal breathing sound was heard once in a certain breathing cycle, we heard the breathing sound over multiple breathing cycles and listened to the abnormal breathing sound multiple times periodically. It is judged as abnormal. Abnormal breathing sounds include continuous raschel sounds and intermittent raschel sounds (hereinafter referred to as continuous rasles sounds, intermittent rasles sounds) that have markedly different characteristics. Sounds and intermittent rales show more detailed features. The doctor makes a judgment based on the characteristics and experience of such breathing sounds.
In order to reproduce the judgment by a doctor using a computer or the like, it is necessary to prepare a large number of cases of abnormal breathing sound samples and normal breathing sound samples. In addition, it is clear that it takes a lot of labor to cut out the acoustic signals of all the samples and give the teacher signals such as “whether breathing sound or non-breathing sound” and “normal breathing sound or abnormal breathing sound”.
As described above, it is desired to provide a technique capable of analyzing the presence / absence of an abnormality from a respiratory sound signal without using a huge amount of learning data.

  In view of such problems, the present invention generates two feature amounts from an input respiratory sound signal by a simple method, and uses the two feature amounts to determine whether or not an abnormal sound is included in the respiratory sound. It is an object of the present invention to provide a respiratory sound analysis device, a respiratory sound analysis method, and a respiratory sound analysis program capable of calculating an evaluation value for determination and an evaluation value for determining the type of abnormal sound.

The respiratory sound analyzer according to the first invention for solving the above-described problem is a peak section that divides an input signal based on a respiratory sound into a plurality of frequency bands and indicates a section having a relatively high level in each frequency band signal. A signal, a non-peak section signal indicating a section other than the peak section, and a peak section synthesized signal obtained by synthesizing the plurality of peak section signals, and a non-peak section synthesized signal obtained by synthesizing the plurality of non-peak section signals. A peak section identifying section to be generated; a feature quantity generating section for generating a first feature quantity from the peak section synthesized signal; and generating a second feature quantity from the non-peak section synthesized signal; and the first feature quantity And a first evaluation value for determining whether or not an abnormal sound is included in the respiratory sound, and a type of the abnormal sound included in the respiratory sound, using the second feature amount Second evaluation value Comprising an evaluation value calculation unit that calculates a.
In the respiratory sound analysis method according to the second invention for solving the above-described problem, an input signal based on a respiratory sound is divided into a plurality of frequency bands, and a peak section indicating a section having a relatively high level in each frequency band signal Generating a signal, a non-peak section signal indicating a section other than the peak section, a peak section synthesized signal obtained by synthesizing a plurality of the peak section signals, and a non-peak section synthesis obtained by synthesizing a plurality of the non-peak section signals. Generating a signal, generating a first feature amount from the peak interval composite signal, generating a second feature amount from the non-peak interval composite signal, the first feature amount, and the second feature amount. And a second evaluation value for determining the type of the abnormal sound included in the respiratory sound, and a first evaluation value for determining whether or not the abnormal sound is included in the respiratory sound. Evaluation value A calculating and.
A respiratory sound analysis program according to a third invention for solving the above-described problem is a peak section that divides an input signal based on a respiratory sound into a plurality of frequency bands, and indicates a section having a relatively high level in each frequency band signal Generating a signal, a non-peak section signal indicating a section other than the peak section, a peak section synthesized signal obtained by synthesizing a plurality of the peak section signals, and a non-peak section synthesis obtained by synthesizing a plurality of the non-peak section signals. Generating a signal, generating a first feature amount from the peak interval composite signal, generating a second feature amount from the non-peak interval composite signal, the first feature amount, and the second feature amount. And a second evaluation value for determining the type of the abnormal sound included in the respiratory sound, and a first evaluation value for determining whether or not the abnormal sound is included in the respiratory sound. To execute the steps of: calculating a value to the computer.

  According to the present invention, two feature values are generated from an input respiratory sound signal by a simple method, and the evaluation for determining whether or not an abnormal sound is included in the respiratory sound using the two feature values. A respiratory sound analysis apparatus, a respiratory sound analysis method, and a respiratory sound analysis program capable of calculating a value and an evaluation value for determining the type of abnormal sound can be provided.

It is a block diagram which shows the respiratory sound analyzer concerning 1st Embodiment of this invention. It is a figure which shows the flowchart of the respiratory sound analyzer concerning 1st Embodiment of this invention. It is a block diagram which shows the peak area identification part of this invention. It is a figure explaining the process of the peak area identification part of this invention. It is a figure explaining the process of the peak area identification part of this invention. It is a figure explaining the process of the feature-value production | generation part of this invention. It is a figure which shows the example of the feature-value which the feature-value production | generation part of this invention produces | generates. It is a figure which shows ratio of each feature-value shown in FIG. It is a block diagram which shows the respiratory sound analyzer concerning 2nd Embodiment of this invention. It is a figure which shows the flowchart of the respiratory sound analyzer concerning 2nd Embodiment of this invention.

Hereinafter, an embodiment of the respiratory sound analyzer of the present invention will be described with reference to the accompanying drawings.
FIG. 1 shows a block diagram of the respiratory sound analyzer of the first embodiment. The respiratory sound analysis device 100 includes a signal acquisition unit 110, a peak section detection unit 120, a feature amount generation unit 130, an evaluation value calculation unit 140, and a determination unit 150. The breathing sound analysis process performed by the breathing sound analyzer 100 will be described with reference to the flowchart of FIG.

  The signal acquisition unit 110 receives a respiratory sound signal collected by a sound collection device (not shown) such as a microphone, performs signal conversion as necessary, and outputs the signal to the peak section identification unit 120. Signal conversion is, for example, A / D conversion.

As shown in FIG. 3, the peak section identification unit 120 includes a plurality of band division filters 2 x (x = 1 to n, the same applies hereinafter), a peak signal generation unit 3 x, and a signal synthesis unit 40. The peak section identification unit 120 receives the input signal S based on the respiratory sound signal in step S201, and divides it into a plurality of (1 to n) frequency bands by the band division filters 21 to 2n. The band division filters 21 to 2n apply the signals s (x) and (s (1) to s (n)) obtained by filtering the input signal S with the bandpass filters respectively included in the corresponding peak signal generation units 31 to 3n. Supply.
In step S202, the peak signal generating unit 3x of the peak section identifying unit 120 determines peak signals ps (x) and (ps (1) indicating peak sections to be described later in the time variation of the frequency from the signal s (x) of each frequency band. ) To ps (n)). The peak signal generation unit 3x supplies the peak signal ps (x) to the signal synthesis unit 40. In the present embodiment, the analysis target frequency that is a target for detecting the peak section is 0 Hz to 2 kHz, but is not limited thereto.

FIG. 4 shows the peak signal ps (n) obtained from the signal s (n) in the frequency band (n) by the peak signal generator 3n. The peak signal ps (n) rises in a section where the level of the signal s (n) in the frequency band is equal to or higher than a certain level indicated by a one-dot chain line (hereinafter, peak section), and the level of the signal s (n) is smaller than a certain level. The signal falls in a section (hereinafter, non-peak section). The peak section identification unit 120 similarly obtains the peak signal ps (x) of the signal s (x) in each divided frequency band, and generates the synthesized signal P shown in FIG. 5 based on the obtained peak signal ps (x). .
It should be noted that for each signal s (x) in the frequency band (x), the reference level is different depending on whether it is a peak interval or a non-peak interval, and the level is not a constant value but a distribution of the signal s (x) You may make it vary based on. For example, the moving average value of the signal s (x) is set as the reference level at each time. In the signal s (x) of the frequency band (x), a section having a relatively high level is a peak section, and a section other than the section identified as the peak section is a non-peak section.

The signal combining unit 40 of the peak section identifying unit 120 generates the combined signal P shown in FIG. 5 from the peak signal ps (x) generated by the peak signal generating unit 3x (step S203). The combined signal P includes a peak section combined signal P1 obtained by combining the peak sections of the respective peak signals ps (x) to be combined, and a non-peak section combined signal P0 obtained by combining the non-peak sections. The synthesized signal P from the time t1 to the time t2 in FIG. 5 is the time t2 from the peak signal ps (2) that first rises at the time t1 among the peak signals ps (1) to ps (n) to be synthesized. The peak section composite signal P1 which is generated by synthesizing the peak signal ps (1) which falls last at, and which rises at time t1 and falls at time t2. The synthesized signal P from the time t2 to the time t3 in FIG. 5 is the time t3 from the peak signal ps (1) falling at the time t2 among the peak signals ps (1) to ps (n) to be synthesized. This is a non-peak interval combined signal P0 generated by combining up to the peak signal ps (n) that rises first.
The peak section synthesized signal P1 of the synthesized signal P is a signal obtained by synthesizing a peak section signal indicating a section having a relatively high level in the signal s (x) of each frequency band (x) obtained by dividing the input signal S. The peak section synthesized signal P0 is a signal obtained by synthesizing a non-peak section signal indicating a non-peak section that is a section other than the peak section in the signal s (x) of each frequency band (x) obtained by dividing the input signal S.
Note that the peak signal ps (x) obtained from the signals s (x) in all frequency bands may be used to generate the synthesized signal P, or the peak obtained from the signal s (x) in the selected frequency band. The signal ps (x) may be used.

An abnormal breathing sound typified by a Russell sound (hereinafter referred to as a “rabble sound”) has a characteristic that an amplitude (or power) is large in a specific frequency component as compared with a normal breathing sound. Therefore, by obtaining the peak signal ps (x) indicating the peak interval and the non-peak interval for the signal s (x) in the frequency band (x), the interval (peak interval) having a high suspicion including abnormal noise and the interval ( The signal s (x) in the frequency band can be roughly divided into (non-peak period).
In addition, the continuous rarity which is an abnormal breathing sound in the present embodiment has a feature that it can be heard continuously for a long time in a narrow frequency band, and the intermittent rarity is heard intermittently for a very short time in a wide frequency band. Due to the characteristics, when the peak section identifying unit 120 divides the input signal S into a plurality of frequency bands (x), the frequency components included in each frequency band may overlap.
The peak section identification unit 120 indicates a peak section signal indicating a peak section having a relatively high level in the signal s (x) of each frequency band from the plurality of signals s (x) obtained by dividing the input signal S into a plurality of frequency bands. A peak signal ps (x) having a non-peak section signal indicating a section other than the peak section, and a non-peak section synthesized signal P1 obtained by synthesizing a plurality of peak section signals and a plurality of non-peak sections. A composite signal P having a section composite signal P0 is generated. The input signal S is identified as a peak interval and a non-peak interval by the synthesized signal P generated by the peak interval identification unit 120.

  The feature value generation unit 130 includes feature values of the peak section and the non-peak section of the input signal S based on the peak section composite signal P1 and the non-peak section composite signal P0 that constitute the composite signal P generated by the peak section identification section 120. For each. The input signal S is converted into a frequency band signal for each unit time frame by using a frequency analysis method such as a short-time fast Fourier transform, and a statistical value for each frequency is obtained in each of a peak period and a non-peak period. calculate. The statistical value is a value indicating frequency distribution characteristics, such as the average value, median value, variance, kurtosis, skewness, difference between the maximum value and minimum value, or the ratio of the power obtained from the amplitude and amplitude for each frequency. It is desirable to use it.

6A and 6B are diagrams showing the frequency distributions obtained by the feature quantity generation unit 130 from the peak and non-peak sections, respectively, with the frequency f on the horizontal axis and the feature quantity of the frequency component on the vertical axis. . In the present embodiment, the frequency distribution obtained from the peak section based on the peak section synthesized signal P1 is the feature quantity CP of the peak section, and the frequency distribution obtained from the non-peak section based on the non-peak section synthesized signal P0 is the feature of the non-peak section. The amount is NP.
In the portion surrounded by the broken line 511 or 512 in the peak section feature amount CP in FIG. 6A, the frequency feature amount is higher than the non-peak section feature amount NP shown in FIG. 6B. It is a part to show, and is a feature part of the feature value CP of the peak section. It shows that many signals with relatively high power were detected at the frequencies f1 and f2.
When the feature quantity NP in the non-peak section in FIG. 6B and the feature quantity CP in the peak section are compared, there is a difference in the distribution of frequency bands surrounded by broken lines 511 and 512. A location where a difference between the feature value CP in the peak section and the feature value NP in the non-peak section is likely to be a difference based on abnormal sound. The presence or absence of abnormal sound is determined by paying attention to such a difference.

The feature amount generated by the feature amount generation unit 130 will be described with reference to FIGS. FIG. 7A shows the feature quantity CP in the peak section and the feature quantity NP in the non-peak section obtained from the input signal S of the respiratory sound including the continuous rales. As described above, the continuous rarity is represented by a signal that is temporally continuous in a narrow frequency band. For example, the continuity sound is a sound like a whistle if it is a high frequency sound, or a sound like a whisper if it is a low frequency sound. In the case where the peak section includes a lot of continuous rales, the feature quantity CP of the peak section indicates that a certain frequency component is included as in a portion surrounded by chain lines 611 and 612. It can be seen that the frequency band surrounded by the chain lines 611 and 612 of the feature quantity CP in the peak section is greatly different from the feature quantity NP in the non-peak section in the same frequency band.
FIG. 7B shows the feature quantity CP in the peak section and the feature quantity NP in the non-peak section obtained from the input signal S of the respiratory sound including intermittent rales. Intermittent rales are represented by signals that are intermittent in time in a wide frequency band as described above. Intermittent rales are sounds such as impulsive noises such as bubble wrap and buzz. It can be seen that when the peak section includes many intermittent radon sections, there is a difference between the feature quantity CP in the peak section and the feature quantity NP in the non-peak section in a wide frequency band as indicated by the chain line 621.

FIG. 7C shows the peak section feature amount CP and the non-peak section feature obtained from the input signal S of the breathing sound that does not include the continuous rarity and intermittent rarity, which are abnormal sounds in the present embodiment. The quantity NP is indicated. In the case of a so-called normal breathing sound that does not include abnormal sounds, a peak interval that occurs due to a difference in amplitude between exhalation and inspiration, ambient noise, or the like is included. The peak signal ps (x) obtained from the input signal S based on a normal breathing sound is detected as a signal having a relatively high power in a specific frequency component or frequency band such as a continuous rarity or intermittent rarity. Not. Accordingly, as shown in FIG. 7C, there is no significant difference between the feature value CP in the peak section and the feature value NP in the non-peak section in the entire frequency to be analyzed.
As described with reference to FIGS. 7A to 7C, the frequency distribution in the peak section and the frequency distribution in the non-peak section are used as feature amounts, respectively, so that the case where the input signal S includes abnormal sound is included. There is a clear difference in the feature values between the peak and non-peak sections. The feature amount of each section may be a value obtained by further calculating statistical values such as variance, kurtosis, and skewness using each element.

The evaluation value calculation unit 140 uses the peak segment feature quantity CP and the non-peak segment feature quantity NP generated by the feature quantity generation unit 130 to determine whether the determination unit 150 includes abnormal sounds in the breathing sound. An evaluation value used for evaluation is calculated (step S205).
The calculation of the evaluation value includes, for example, a value indicating vector similarity such as cosine similarity and cross-correlation value, vector distance such as Euclidean distance and Mahalanobis distance, feature quantity CP in the peak section and feature quantity NP in the non-peak section. Statistical values (average value, variance, etc.) based on the ratio or difference of each element can be used. When the vector size is large, it may be difficult to display the similarity feature. Therefore, it is desirable to reduce the vector size while preserving the vector feature by, for example, combining adjacent elements into one element by averaging. Further, the feature amount may be normalized according to the characteristic of the evaluation value. The vector is a feature quantity CP and a feature quantity NP, and is a set of feature quantities (scalars) at each frequency f.

The determination unit 150 determines whether or not the evaluation value obtained by the evaluation value calculation unit 140 satisfies the condition (step S206). If the condition is not satisfied, it is determined that there is no abnormal sound (step S207), and if the condition is satisfied, it is determined that there is an abnormal sound (step S208).
When the cosine similarity is used to calculate the evaluation value, the higher the similarity between the feature amounts CP and NP, the closer to 1. Therefore, a value close to 1, for example, 0.9 is set as a threshold value, and the condition is whether the similarity is equal to or less than the threshold value. When the Euclidean distance is used to calculate the evaluation value, the distance between the vectors varies depending on the normalization condition, but the value becomes closer to 0 as the similarity between the feature quantity CP and the feature quantity NP increases. Therefore, a threshold value greater than at least 0 is set according to the normalization condition, and the condition is whether the similarity is equal to or greater than the threshold value.
When the ratio of each element included in the feature value is used for calculating the evaluation value, the higher the similarity between the feature value CP and the feature value NP, the more elements having a value closer to 1.

8A to 8C show the ratio R of the feature value CP to the feature value NP obtained for each frequency between the feature value CP and the feature value NP shown in FIGS. 7A to 7C, respectively. As shown in FIG. 8A, the feature value CP and the feature value NP of the respiratory sound including the continuous rales shown in FIG. 7A have a ratio R of 1 larger in some frequency bands 711 and 712. Exceed. The feature value CP and feature value NP of the breathing sound including intermittent rales shown in FIG. 7B have a ratio R exceeding 1 in a wide frequency band 721 as shown in FIG.
As is clear from FIGS. 8A and 8B, the frequency band 721 in which the ratio R obtained from the breathing sound including the intermittent sound is greater than 1 is the ratio R obtained from the breathing sound including the continuous sound. Is wider than the frequency bands 711 and 712 above 1. The frequency bands 711 and 712 in which the ratio R obtained from respiratory sounds including continuous rales shown in FIG. 8A exceeds 1 are narrow including a very small number of frequencies with respect to all frequencies used in the analysis processing. It is an area. A frequency band 721 in which the ratio R obtained from respiratory sounds including intermittent rales shown in FIG. 8B exceeds 1 is a wide region including almost all frequencies used for analysis processing.

  As shown in FIG. 8C, the feature value CP and the feature value NP of the respiratory sound that do not include the abnormal sound shown in FIG. About 1 is shown. As conditions based on characteristics that differ depending on the presence or absence of abnormal sounds described in FIG. 8A to FIG. 8C, the average value of the ratio of each element is greater than 1 (for example, 1.5 or more), or a plurality of conditions And a frequency sufficiently higher than 1 (for example, 2 or more). Further, dispersion, kurtosis, and skewness may be used.

  In step S209, display information is generated based on either the determination that the abnormal sound is not included in step S207 or the determination that the abnormal sound is included in step S208, and the analysis process of the respiratory sound is terminated. With the respiratory sound analysis processing of the first embodiment, it is possible to determine whether or not an abnormal sound is included in the respiratory sound by a simple method without requiring data for learning.

Second Embodiment
In the respiratory sound analysis process of this embodiment, when it is determined by the respiratory sound analysis process described in the first embodiment that an abnormal sound is included in the respiratory sound, the type of the abnormal sound is further identified. In this embodiment, a continuous ra sound and an intermittent ra sound are identified.
The respiratory sound analysis apparatus 101 of the present embodiment shown in FIG. 9 is different from the respiratory sound analysis apparatus 100 of the first embodiment in that an evaluation value calculation unit 141 and a determination unit 151 are provided. Since the configuration other than the evaluation value calculation unit 141 and the determination unit 151 included in the respiratory sound analysis device 101 of the present embodiment is the same as that of the respiratory sound analysis device 100 of the first embodiment, the same reference numerals are given and the description is omitted. . Respiratory sound analysis processing performed by the respiratory sound analysis apparatus 101 will be described with reference to the flowchart of FIG.

  The processing from when the input signal S is input to the respiratory sound analyzer 101 until the feature amount generation unit 130 generates the feature amount is the same as Step S201 to Step S204 in the first embodiment, and thus is illustrated in FIG. In the flowchart, the same reference numerals are given, and description thereof is omitted.

The evaluation value calculation unit 141 determines not only an evaluation value for determining the presence or absence of an abnormal sound, but also whether or not the abnormal sound includes a continuous rarity and whether or not the abnormal sound includes an intermittent rarity. An evaluation value is also calculated. In step S801, the evaluation value calculation unit 141 obtains an evaluation value as follows.
As shown in FIG. 7 (A), the feature value CP of the breathing sound including continuous rales includes a peak in a part of the frequency band indicated by the chain line 611 and the chain line 612. In many cases, the change includes a large part. On the other hand, as shown in FIG. 7B, the feature value CP of the respiratory sound including the intermittent sound is a case where the change in the feature value between adjacent frequencies includes the continuous sound shown in FIG. Often it is more lenient. When the differential value of the feature value CP is calculated based on such features, in the case of the respiratory sound including the continuous rales shown in FIG. 7A, the differential value has a sharp change, but FIG. In the case of respiratory sounds including intermittent rales shown in (2), the differential value has a constant value or a gradual change. Therefore, in the present embodiment, a second-order differential value is obtained, and the number of zero crossings is used as an evaluation value as to whether or not continuity rales are included.
In addition, as other evaluation values, statistical values such as dispersion indicating variation in the distribution of differential values, kurtosis indicating the degree of kurtosis, etc. are used, or respiratory sound feature values shown in FIGS. A statistical value may be calculated and used from the ratio R obtained from the CP and the feature quantity NP.

  As shown in FIG. 8B, the ratio R obtained from the breathing sound including intermittent rarity exceeds 1 in almost all frequencies used in the analysis processing as described in the first embodiment. . On the other hand, the ratio R obtained from the respiratory sound including the continuous rales shown in FIG. 8 (A) has a frequency exceeding 1 less than that of the intermittent rales. Therefore, in this embodiment, the threshold value is set to a value of 1 to 2, and the value calculated for the ratio of the frequency exceeding the threshold value to the entire frequency used for the analysis is an evaluation value as to whether or not intermittent rales are included. did.

  In the present embodiment, the determination unit 151 first performs a determination process on whether or not an abnormal sound is included in the respiratory sound in step S802 using the evaluation value calculated by the evaluation value calculation unit 141 in step S801. Since the determination process is the same as that in steps S206 to S208 performed by the determination unit 150 of the first embodiment, the description of steps S802 to S804 is omitted.

If it is determined in step S804 that there is an abnormal sound, the process proceeds to step S805, and the determination unit 151 determines whether or not the evaluation value for evaluating whether or not the continuity rales are included satisfies a predetermined condition. . If it is determined that the condition is satisfied, the breathing sound includes a continuous rarity (step S807). If it is determined that the condition is not satisfied, the breathing sound does not include a continuous rarity (step S806). Here, the determination condition in step S805 is whether the number of zero crossings of the second-order differential value is three or more.
In a succeeding step S808, it is determined whether or not an evaluation value for evaluating whether or not an intermittent sound is included satisfies a predetermined condition. If it is determined that the condition is satisfied, the breathing sound includes an intermittent sound (step S810). If it is determined that the condition is not satisfied, the breathing sound does not include an intermittent sound (step S809). Here, the determination condition in step S808 is whether or not the ratio of the frequency exceeding the threshold to all frequencies used in the analysis is 60% or more.

In step S811, display information is generated and analyzed for the presence or absence of an abnormal sound, and if there is an abnormal sound, whether the abnormal sound includes a continuous rabble or whether the abnormal sound includes an intermittent rabble. The process ends.
With the respiratory sound analysis processing of the second embodiment, it is possible to determine whether or not an abnormal sound is included in the respiratory sound by a simple method without requiring data for learning. , Intermittent rales) can also be determined.

Note that the respiratory sound analyzers of the first and second embodiments may not include the determination units 150 and 151. The feature amount and the evaluation value may be displayed on the display unit 200, and the user may determine whether or not the abnormal sound is included based on the displayed information and determine the type of the abnormal sound.
In addition, you may make a computer perform the respiratory sound analysis process which the respiratory sound analyzer of 1st Embodiment and 2nd Embodiment performs as a program.

DESCRIPTION OF SYMBOLS 100, 101 Respiration sound analyzer 110 Signal acquisition part 120 Peak area identification part 130 Feature-value production | generation part 140,141 Evaluation value calculation part 150,151 Judgment part 200 Display part

Claims (4)

The input signal based on the respiratory sound is divided into a plurality of frequency bands, and a peak section signal indicating a relatively high level section in each frequency band signal and a non-peak section signal indicating a section other than the peak section are generated. A peak section identifying signal for generating a peak section synthesized signal obtained by synthesizing a plurality of the peak section signals, and a non-peak section synthesized signal obtained by synthesizing the plurality of non-peak section signals;
A feature quantity generating unit that generates a first feature quantity from the peak section synthesized signal and a second feature quantity from the non-peak section synthesized signal;
Using the first feature value and the second feature value, a first evaluation value for determining whether or not the respiratory sound includes an abnormal sound, and the abnormal sound included in the respiratory sound A respiratory sound analyzer comprising: an evaluation value calculation unit that calculates a second evaluation value for determining the type of the respiratory sound.   Based on the first evaluation value, it is determined whether or not the abnormal sound is included in the breathing sound, and the type of the abnormal sound is determined based on the second evaluation value. The respiratory sound analyzer according to claim 1, further comprising a determination unit. The input signal based on the breathing sound is divided into a plurality of frequency bands, and a peak section signal indicating a relatively high level section in each frequency band signal and a non-peak section signal indicating a section other than the peak section are generated. Steps,
Generating a peak interval synthesized signal obtained by synthesizing a plurality of peak interval signals, and a non-peak interval synthesized signal obtained by synthesizing a plurality of non-peak interval signals;
Generating a first feature quantity from the peak section synthesized signal and generating a second feature quantity from the non-peak section synthesized signal;
Using the first feature value and the second feature value, a first evaluation value for determining whether or not the respiratory sound includes an abnormal sound, and the abnormal sound included in the respiratory sound Calculating a second evaluation value for determining the type of breathing sound. The input signal based on the breathing sound is divided into a plurality of frequency bands, and a peak section signal indicating a relatively high level section in each frequency band signal and a non-peak section signal indicating a section other than the peak section are generated. Steps,
Generating a peak interval synthesized signal obtained by synthesizing a plurality of peak interval signals, and a non-peak interval synthesized signal obtained by synthesizing a plurality of non-peak interval signals;
Generating a first feature quantity from the peak section synthesized signal and generating a second feature quantity from the non-peak section synthesized signal;
Using the first feature value and the second feature value, a first evaluation value for determining whether or not the respiratory sound includes an abnormal sound, and the abnormal sound included in the respiratory sound And a step of calculating a second evaluation value for determining the type of the respiratory sound analysis program.

Images