JP3780273B2 - Body reaction information detection system while sleeping - Google Patents

Description

  The present invention relates to a system for detecting biological information from sounds of a respiratory system that a person emits while sleeping.

Traditionally, sleeping habits and sleep are considered to be related to the condition of the person's body, and studies on wrinkles have been conducted.
For example, it is said that many people who are prone to scratching are often obese, or even the same person is struggling when tired or hard to sleep. Not only that, but people with respiratory and nervous system illnesses can develop terribly large epilepsy, and attempts to discover hidden serious diseases and signs of seizures from sputum measurements are also made. ing.

  In addition, unconscious body reaction information is likely to appear at bedtime. This is because the unconscious body reaction information is hidden at the time of awakening because of the conscious body control by the cerebral function. The above conscious body control has a much larger dynamic range such as the range and speed of movement than the unconscious body reaction, so the unconscious body reaction information is hidden by such a large change at awakening. There is a possibility that. In contrast, during sleep, cerebral function is paused or reduced, so an unconscious physical reaction should be likely to appear. Such unconscious body reactions are thought to contain essential information of the body that is medically valuable.

For example, body reaction information related to important physical problems related to health may be hidden by a conscious reaction due to cerebral function when awakened. In order to detect such unconscious body reaction information, research has been conducted on sneezing and sleeping.
And in order to study what kind of hemorrhoids are related to what kind of disease, we collect actual hemorrhoids, analyze it, and what kind of hemorrhoids the subject uses Attempts have been made to find out. Specifically, a person hears and judges a roaring sound recorded using a microphone and performs frequency analysis to classify the type of the roar.
Since such a system is actually conducted, no prior art literature search is conducted.

However, if a microphone is placed near the patient's bed to collect stuttering, but the face of the subject who is sleeping is facing away from the microphone, the microphone In some cases, the noise that can be captured by the computer becomes smaller, and the necessary data cannot be collected.
In other words, when the subject's face is facing away from the microphone, the recorded sound of 鼾 decreases, and when the face of the subject faces the microphone, the recorded sound of 鼾 increases. The strength of the sound cannot be judged. Also, during bedtime, the sound of turning over and the sound of bedclothes are generated, and these are recorded together with the sound of the kite. It may be worn out.

  On the other hand, it is known that stuttering includes a normal conversation voice and high frequency components that are hardly included in other sounds. For this reason, in order to remove noise other than stuttering from the recorded audio data, it may be possible to leave only the high frequency region and delete the audio data in other frequency amount regions. However, the high-frequency component is a small percentage of the total stuttering and, as described above, the high-frequency component is extracted when the level of sound that can be recorded becomes low depending on the orientation of the subject's face relative to the microphone. It becomes difficult to do. In particular, since high-frequency sound is easily absorbed by bedding and the like, it has been practically impossible to collect high-frequency components with the microphone when the face is away from the microphone.

  Also, monitor the subject who changes the direction of the face, such as by turning over, and move the microphone according to the movement, or install many microphones around the face, Although it is conceivable to use a device that allows a person to record a song regardless of the direction, a large-scale device is required. When a large-scale device is used, not only does the device cost increase, but the subject surrounded by the oversized device may be tense and unable to sleep. In addition, when a subject is put on a microphone, the subject often cannot sleep.

  An object of the present invention is a system capable of accurately collecting body reaction information from respiratory sounds such as breathing sounds and stuttering of a subject who is sleeping in a state in which the subject is not stressed. Is to provide.

According to a first aspect of the present invention, there is provided a sound capturing means for capturing a sound of a respiratory system while sleeping, a digital data converting means for converting the sound captured by the sound capturing means into digital data, and data for analyzing the digital data. A processing unit, wherein the data processing unit sets a processing unit for the digital data and applies an algorithm for calculating the first Lyapunov exponent for each processing unit to calculate the stability of fluctuation, The feature is that it has a function of outputting it.

The sound of the respiratory system is a sound emitted from the respiratory tract from the nose or mouth to the lungs, such as a subject's sleep or phlegm.

According to a second aspect of the present invention, there is provided a sound capturing means for capturing sound of a respiratory system during sleep, a digital data converting means for converting the sound captured by the sound capturing means into digital data, and data for analyzing the digital data. The data processing means analyzes the digital data by an extended chaotic method extended to respond to changes in the physical state during sleep, and fluctuates from the brain function index defined by this extended chaos theory. It has a feature in that it has a function of calculating the stability of and outputting it.

The third invention is based on the first or second invention, and comprises data storage means for storing in advance a reference level of fluctuation stability, a reference pattern of fluctuation stability over time, and the like. The processing means compares the stability of fluctuation calculated based on the sound input from the sound capturing means and the reference level or reference pattern stored in the data storage means to determine a change in the physical condition. It is characterized in that

According to the first to third inventions, it is possible to detect necessary body reaction information without applying stress due to measurement to a subject.

In particular, in the second invention, it becomes possible to grasp in real time a fine change in the stability of fluctuation. Therefore, it is possible to predict sleep disorder and prevent it.
Further , according to the third invention, it is possible to automatically determine the body reaction information while sleeping.

A first embodiment of the present invention will be described with reference to FIGS.
In this system, a microphone 2 which is a sound capturing means for capturing a living body sound during sleep is connected to a computer 1. The microphone 2 is a bedside of the subject's bed, and is fixed at a position where it does not get in the way when the subject turns over.

  The computer 1 stores digital data conversion means 3 for converting sound input from the microphone 2 into digital sound data, and sound data storage for storing the digital sound data converted by the digital data conversion means 3. Means 4 and data processing means 5 for analyzing the sound data are provided. A program storage unit 6 is connected to the data processing unit 5, and the sound data is analyzed by an analysis program stored in the program storage unit 6.

Further, a display 7 is connected to the data processing means 5 as data output means.
Note that the sound data storage means 4 for storing the sound data is not provided, and raw data storage means for storing the recorded sound before digital conversion is provided between the microphone 2 and the digital data conversion means 3. Also good. Further, an integrated sound capturing means and digital data converting means may be used.

Hereinafter, a procedure for collecting biological sounds while the subject is sleeping using the system shown in FIG. 1 will be described.
First, when the computer 1 and the microphone 2 are turned on, the subject's sleep or stuttering is input to the computer 1 by the microphone 2.
In the computer 1, digital sound data is generated by the digital data conversion means 3. This digital sound data is stored in the sound data storage means 4 and further analyzed by the data processing means 5.

  The data processing means 5 applies an algorithm for calculating the first Lyapunov exponent and calculates the stability of fluctuation of the sound data. The first Lyapunov exponent is an index representing the spread of attractors created from time-series data having chaotic properties. In the chaos theory for calculating the first Lyapunov exponent, it is assumed that the dynamics are stable. That is, in the first embodiment, it must be assumed that the state of the body that emits the sound data is stable.

  However, the dynamics in this first embodiment, that is, the physical state of the subject who is sleeping is not as stable as assumed by the chaos theory, and sound data such as sleep that a person emits is originally converged data. Absent. Therefore, strictly speaking, the first Lyapunov exponent cannot be calculated from the sound data. Therefore, in the first embodiment, the sound data over time is divided by a preset processing unit, and the result calculated by the same method as the first Lyapunov exponent in the unit range is used for the fluctuation of the present invention. Stability is assumed.

  The first Lyapunov exponent indicates the degree of fluctuation of chaotic time-series data. However, the sound data collected during sleep is not completely chaotic, and the data includes various types of noise that are not chaotic. In nature, breathing during sleep is periodic, and the behavior may be considered as chaos, but in detail, the stimulation to the brain given by the external environment such as noise and temperature, the physical condition of the subject, etc. It is considered that the body condition based on the brain function is reflected in the sound data.

On the other hand, when the first Lyapunov exponent is calculated from data including noise, it is known that the value is calculated larger than when there is no noise. Therefore, in the system of the first embodiment, the value calculated using the first Lyapunov exponent calculation algorithm is not the chaos index but the fluctuation stability. The larger the stability value of this fluctuation is, the more unstable the fluctuation is.
Therefore, by calculating the stability of the fluctuation, it is possible to estimate the load on the brain as the body reaction information of the subject at that time.
When the value of stability is large, a lot of noise is included in chaotic fluctuation, and it is determined that the brain is loaded.

The experimental results for verifying this will be described next.
Graph A in FIG. 2 shows the result of calculating the value of the stability of fluctuation by analyzing the sleep and stuttering captured by the microphone 2 from the sleeping subject as described above. However, in this experiment, the movement of the subject and the sound in the room are monitored in real time in a monitoring room different from the room where the subject is sleeping, using an infrared camera and a microphone. Then, the movement of the subject and the state of wrinkles are recorded so that the state of the subject can be associated with the value of the stability of fluctuation.

The experimental results shown in FIG. 2 are the calculation results when the sampling frequency is 44.1 kHz, and the 6-minute moving average is calculated with the processing unit of the present invention as 1 second. That is, the calculation of the stability value of the fluctuation and the averaging process with a width of 6 minutes are performed with a shift of 1 second.
As described above, when processing is performed in a processing unit much longer than the sampling period, the calculated stable value may change depending on the processing timing. Therefore, the average value is calculated over a long period of 6 minutes, and the tendency of the change is observed.

In graph A of FIG. 2, the first period (1) in which the stability value continues to be low, the second period (2) in which the stability value gradually increases, and a slightly higher state is stable. It is divided into the third period (3) and the fourth period (4), which has risen sharply.
The state of the subject corresponding to the changes in these graphs did not change in the first period (1) and the second period (2), and the sleep was quiet and the patient seemed to sleep. Looked. However, as soon as the third period (3) was entered, sputum and apnea were repeated, and in the fourth period (4), severe sputum continued. In other words, the second period (2) and the third period (3) are divided not by the calculated stability value but by the state of the subject.

From the above, it can be seen that the value of the stability of fluctuation is likely to be related to the state of the subject sleeping.
Actually, as in the fourth period (4), when the value of the stability becomes very high, the player is stroking intensely. In this case, intense drought refers to drought that occurs under intense breathing and has nothing to do with the magnitude of the sound at the recording level. That is, by calculating the value of fluctuation, the sleeping state can be estimated to some extent even if it is not visually observed. Moreover, it is possible to detect the physical condition regardless of the recording level of sound data, that is, the size of recording, as in the conventional collection of stuttering.

On the other hand, in the third period (3), apnea and sputum actually occur alternately, which is clearly different from the second period (2). However, the stability value is almost the same in the second half of the second period (2) and the stability value in the third period (3), so the change in the third period (3) is the stability data. It is difficult to understand only from it.
As described above, the state of the third period (3) cannot be read from the graph of FIG. 2 for the following reason. The fluctuation stability in the first embodiment is calculated by determining the processing unit, and is an average value over a long period of 6 minutes. Therefore, the graph of FIG. Change does not appear . For this reason , it is considered that the state of the third period does not appear clearly in this graph .

However, if the same experiment is repeated and the change in stability is associated with the observation of the subject, the stability value is low and stable as in the first period (1) of graph A. It is also possible to predict that the breathing state has changed when the value increases compared to when the person is.
Regardless of the recording level of the breathing sound, it can be said that if the person is striking hard, the stability value is high and the load on the brain is large. If such a state continues, it can be assumed that the body is not resting even if the bedtime is long.

FIG. 3 shows the result of calculating the stability of fluctuation from the breathing sound while sleeping by the system of the second embodiment.
The system of the second embodiment has the same overall configuration as the system of the first embodiment shown in FIG. Therefore, FIG. 1 is also used to explain the second embodiment.
However, the analysis program stored in the program storage means 6 is different from that in the first embodiment.
Also in the second embodiment, the method for taking in breath sounds such as sleep and snoring from the microphone 2 and taking the digitized data into the sound data storage means 4 is the same as in the first embodiment.

The analysis program of the second embodiment is a program that calculates a brain function index, which will be described later, from sound data stored in the sound data storage means 4, and outputs this as a fluctuation stability value. .
The brain function index is an index defined by an extended chaos theory that is originally extended from a chaos theory that does not assume a phenomenon in which dynamics change, to cope with a phenomenon in which dynamics change.
The extended chaos theory is defined in Japanese Patent Application No. 2003-045386 filed by the inventors of the present invention.

When the brain function index is SiCECA (s (t)) and the sound data is s (t), the brain function index is given by Equation 1.

T (t) in the above equation 1 is a value corresponding to the dynamics of the utterance, but it may be considered constant when sleeping. However, it may be classified according to the posture of the subject while sleeping.
In addition, ε _S in the above equation 1 is the Siseka neighborhood distance given by equation 2.

In addition, the brain function index and the Shiseka vicinity distance are also described in detail in the Japanese Patent Application No. 2003-045386.
In the system of the second embodiment, the data processing means 5 calculates and outputs the brain function index based on the collected sound data. At that time, for all data sampled at the sampling frequency of 44.1 kHz, the above Siseka vicinity distance is calculated to calculate the brain function index. Graph B in FIG. 3 shows the result of averaging the calculated brain function index with a moving average width of 30 seconds and a moving interval of 1 second.

The experimental results in FIG. 3 are obtained by analyzing sound data collected from the same subject as in the first embodiment.
Graph B in FIG. 3 is also a graph showing the stability of fluctuation calculated based on sound data, similar to graph A of the experimental results of the first embodiment shown in FIG. 2, and is divided into four periods. Yes. The external appearance of the subject in each period is the same as in the first example.
However, the graph B of the second embodiment shows a fine change in the stability of fluctuation compared to the graph A of the first embodiment. According to this graph B, compared to the first period (1), the second period (2), which is higher than the first period (1), and the overall level is not much different from the second period (2), but the fluctuation is severe The third period (3) can be clearly distinguished from the fourth period (4), which is very high and changes rapidly.

In particular, in the graph A in FIG. 2, the second period (2) and the third period (3), which were difficult to distinguish, can be clearly distinguished, and the third period (3) and the state in which apnea has occurred. Can be associated.
Thus, the graph B shows a fine change compared to the graph A. The reason for this was that all the sampled data was analyzed by the calculation of the data processing means 5 in the second embodiment, and the average processing was performed with a small width.
Here, since it is an experiment, the sound data stored in the sound data storage unit 4 is analyzed, but in practice, data processing may be performed while collecting respiratory sounds. If data processing is performed while collecting respiratory sounds, meaningful data can be obtained in real time, so body reaction information is detected from the change in stability of fluctuations above, and the previous state is predicted You can also

For example, in the second period (2), in the visual observation with respect to the subject, the sleep is quiet and is not different from the first period (1). However, as shown in FIG. 3, in the second period (2), the stability of fluctuation based on the brain function index is different from that in the first period (1) in both level and fluctuation range.
The brain function index is considered to be noise included in the fluctuation of sound data. Such noise is included because the subject's brain is stimulated and a signal other than controlling breathing during sleep is output. In other words, even if it looks as if it is a good night's sleep, it is considered that the brain is more stressed in the second period (2) than in the first period (1).
If it is known that there is a high possibility that apnea will occur after such a load has been applied for a certain period of time, the state of the third period (3) in which sputum and apnea are repeated, Prediction can also be made in the second stage (2).

If the stability of fluctuation is calculated as in the first and second embodiments, noise that disturbs the chaotic qualitative nature of the living body can be detected regardless of the volume level of the recorded data. As a result, it is possible to distinguish between sleep at rest and abnormal sputum and to detect body reaction information.
As in the past, when frequency analysis of captured sound data, the magnitude of the sound affects the analysis result, so it was difficult to collect body reaction information from the sleeping sound while sleeping, According to the system of the present invention, body reaction information can be detected without applying stress to the subject.

  In addition, by calculating the stability of fluctuation and not only distinguishing sleep and mania from the level, but also calculating the stability of fluctuation of sleep at rest of each subject in advance, By contrasting, it is possible to more accurately predict the occurrence of abnormalities during sleep of a specific individual. Therefore, it becomes possible to take preventive measures against the abnormality. For example, if the patient wakes up immediately before entering the apnea state or gives a stimulus by shaking the bed, the apnea state can be stopped.

Furthermore, it is conceivable to determine an abnormality during sleeping according to not only the absolute value of the stability of fluctuation itself but also a change pattern of the stability of fluctuation.
Accordingly, if the reference level of fluctuation stability corresponding to an abnormality such as a disease, the reference change pattern, and the like are stored in advance in the data storage means provided in association with the data processing means 5, the data processing section 5 However, it becomes possible to quickly find an abnormality by comparing the data of the subject who is sleeping and the above-mentioned standard.

In addition, sleep disorders such as apnea and severe hemorrhoids are often not noticed by the person, and if there is a sleep disorder, physical fatigue can be removed even though the patient has slept for a long time. There may be no sleep, or you may be afraid of sleepiness during work. However, in the system of the present invention, the sleep disorder as described above can be diagnosed without applying stress to the subject. Therefore, it is particularly effective in determining the suitability and health management of people in occupations that save lives, such as pilots and train drivers.

1 is a system configuration diagram of an embodiment of the present invention. It is the graph which showed stability of fluctuation of sound data of the 1st example. It is the graph which showed the stability of fluctuation of the sound data of the 2nd example.

Explanation of symbols

1 Computer 2 Microphone 3 Digital Data Conversion Unit 4 Sound Data Storage Unit 5 Data Processing Unit 7 Display

Claims (3)

A sound capturing means for capturing the sound of the respiratory system while sleeping; a digital data converting means for converting the sound captured by the sound capturing means into digital data; and a data processing means for analyzing the digital data, The data processing means sets a processing unit for the digital data , applies a algorithm for calculating the first Lyapunov exponent for each processing unit, calculates the stability of fluctuation, and outputs the function A body reaction information detection system while sleeping. A sound capturing means for capturing the sound of the respiratory system while sleeping; a digital data converting means for converting the sound captured by the sound capturing means into digital data; and a data processing means for analyzing the digital data, The data processing means analyzes digital data using an extended chaos theory method extended to correspond to changes in the physical state during sleep, and calculates the stability of fluctuation from the brain function index defined by this extended chaos theory. , A body reaction information detection system while sleeping, with a function to output it. The data storage means is provided with pre-stored reference levels for fluctuation stability, reference patterns of fluctuation stability over time, and the data processing means is calculated based on the sound input from the sound capturing means. The body reaction information detection system during sleep according to claim 1 or 2, wherein a change in physical condition is determined by comparing the stability of fluctuation with a reference level or a reference pattern stored in the data storage means .

Images