WO2013061415A1 - Respiration measurement system and rem sleep assessment system - Google Patents

Abstract

The invention measures respiratory rate with high precision even with sensing data from a wristband sensor node. The respiration measurement system is provided with: a sensor that is worn on the arm and acquires sensing data representing movements of the arm; a main cyclic component-detecting unit that periodically tabulates sensing data for a prescribed elapsed period and detects the main cyclic component thereof; a respiratory information acquisition validity-assessing unit that assesses whether or not said main cyclic component is valid as a respiration-dependent cyclic component on the basis of the frequency and/or amplitude of the main cyclic component according to previously determined criteria; and a storage unit that stores the frequency of the main cyclic component as a respiratory rate correlated with time information when the main cyclic component is determined to be valid as a respiration-dependent cyclic component.

Description

Respiration measurement system and REM sleep determination system

The present invention relates to a respiration measurement system and a REM sleep determination system, and more particularly to a respiration measurement system and a REM sleep determination system that measure the sleep and / or respiration state of a subject for a long period of time.

The frequency of breathing during sleep is closely related to the depth of sleep and the presence or absence of disease. Therefore, long-term measurement of breathing during sleep is an effective means for health management and early detection of diseases.
Conventionally, as means for measuring the respiratory frequency during sleep, an acceleration sensor, a strain gauge, or an expiration sensor that detects airflow in the nostril are known. For example, in Patent Document 1, the respiratory motion of the chest is measured by a triaxial acceleration sensor attached to the chest, the output value is scalarized, the low frequency component is extracted by a low-pass filter, the peak is detected, and the peak for one minute is detected. A respiration rate estimation technique for outputting a number as a respiration frequency is disclosed.

The technology disclosed above is premised on an acceleration sensor to be worn on the chest. Therefore, it was unsuitable for long-term measurement at home. This is because the chest sensor has the following problems.
・ The chest sensor has a feeling of pressure on the chest when worn, and the sleep state that is different from everyday life is measured.
・ It is important that the chest sensor is properly positioned, and it is difficult to perform correct measurement without expert supervision, making it difficult to measure at home.
-The burden on the chest sensor is 24 hours. Therefore, it is usually necessary to remove it and wear it consciously before going to bed. For this reason, for example, when you go to sleep while watching TV or drink and go to sleep, there is a possibility that data that is biased in the situation leading to sleep may be recorded, not remaining in the measurement record .
On the other hand, bracelet-type acceleration sensors that are worn on the wrist are known as means for measuring the sleep state in daily life. The bracelet-type acceleration sensor is less expensive and less expensive, and it is the same as a normal wristwatch, so it is suitable for long-term continuous measurements at home. Non-Patent Document 1 points out that a bracelet type acceleration sensor can effectively distinguish between sleep and awakening.

Japanese Patent No. 3809847 JP 2008-210363 A JP 2011-94881 A

However, it has not been possible to measure the respiration frequency using a bracelet-type acceleration sensor.
Conventionally, no means has been known for measuring the respiratory frequency during sleep, for example, with less burden on the subject using the bracelet type acceleration sensor as described above. However, the present inventors have discovered that the low-frequency component of the acceleration data measured by the wristband type acceleration sensor may reflect the periodic movement of the trunk derived from respiratory motion. FIG. 21a shows the frequency spectrum of the scalar value of the acceleration waveform in the actual sleep data of less than 7 minutes. In this example, there is a strong peak at 0.22 Hz (corresponding to a breathing frequency of 13.2 times per minute). The case is shown.

This phenomenon occurs when the wrist on which the bracelet type acceleration sensor 1 is mounted is in a position where it is affected by the respiratory motion of the trunk, as shown in FIG. 22a. Therefore, for example, when the wrist is separated from the trunk as shown in FIG. 22b, the periodicity of respiration does not appear in the acceleration data as shown in FIG. 21b.

For this reason, the method disclosed in Patent Document 1 cannot be applied to the acceleration data of the arm. Acceleration data obtained from the chest reflects the breathing motion regardless of the user's body posture and limb position. Therefore, as disclosed in Patent Document 1, it is possible to accurately estimate the respiration frequency simply by finding main frequency components from a specific frequency region. However, in the data measured from the arm, it is rare that the breathing motion is reflected in the acceleration. For this reason, in many cases, the influence of noise (temperature change, sensor measurement noise, etc.) on the sensor value is dominant, and simply finding the main frequency component has no meaning not related to breathing. The value will be calculated.

This is illustrated by example in FIG. FIG. 23a shows the respiration frequency data for every minute measured by the nostril breath sensor, and FIG. 23b shows the data of the bracelet type acceleration sensor worn on the wrist during the same period as the breath sensor. 1 is an estimated respiration frequency per minute calculated when the method disclosed in 1 is applied. Although there are places where the same value as the actual measurement data is calculated, there are many completely different values, and it is difficult to know the tendency of the subject's breathing during sleep from this estimation result itself. FIG. 24a shows the result of calculating the measured respiratory frequency and the estimated respiratory frequency per minute by simultaneously measuring the breath sensor and the wristband type acceleration sensor for 21 people each night. In this scatter diagram, it can be seen that the estimated respiratory frequency and the measured respiratory frequency vary greatly from each other.

The present invention has been made in view of the above, and a respiration measurement system capable of measuring respiration frequency with high accuracy even with sensing data from a wristband type sensor node, and REM sleep based on respiration frequency (REM An object of the present invention is to provide a REM sleep determination system for determining (sleep).

The present invention includes a plurality of means for solving the above-described problems. If one example (first solving means) is given,
A sensor that is attached to the arm and acquires sensing data indicating the movement of the arm;
A main period component detector that periodically collects sensing data of a predetermined period in the past and detects a main period component that is the main period;
Respiration information acquisition success determination unit that determines whether the main period component is effective as a period component depending on respiration according to a predetermined determination condition regarding the magnitude or frequency of the main period component;
There is provided a respiration measurement system including an accumulating unit that accumulates the frequency of the main period component as a respiration frequency in association with time information when it is determined to be effective as a period component due to respiration.

With such a configuration, it is possible to measure the respiration frequency with high accuracy even with sensing data from the wristband type sensor node.
As an example, a respiration rate measurement system that can measure respiration frequency with high accuracy even with a wristband type accelerometer is provided by excluding the case where the breathing motion to be sensed is not reflected in the acceleration data as a missing value. To do.

For example, the following problems can be solved for the above problem. The “peak score” that evaluates the peak intensity in the spectrum of the extracted main frequency component is defined as follows (Equation 1):

Respiratory motion may not be reflected by adopting the detected peak only when the peak score value exceeds the threshold value (ie, it was a sufficiently strong peak), otherwise it is not detected. Respiratory frequency can be accurately estimated from acceleration data obtained from an existing acceleration sensor attached to the wrist. According to this method, as illustrated in FIG. 23c, although there are many missing values, it is possible to estimate the respiration frequency that follows the measured values with high accuracy. FIG. 24b also shows that the respiration frequency estimated by the proposed method can closely approximate the actually measured respiration frequency even in the data of 21 persons.

According to the second solution of the present invention,
A respiratory measurement system as described above;
A sleep detector that detects sleep from the sensing data;
Basal respiration frequency calculation for calculating a basal respiration frequency for every elapsed time after bedtime based on a respiration frequency for a predetermined period accumulated in the accumulation unit of the respiration measurement system and time information. And
A divergence calculating unit that calculates a divergence between the calculated basic respiration frequency and a respiration frequency that is actually measured and accumulated in the accumulation unit;
A REM sleep determination unit that determines REM sleep from the difference between the basal respiration frequency and the actually measured respiration frequency;
There is provided a REM sleep determination system including a second storage unit that stores time information of the determined REM sleep.
With such a configuration, the REM sleep period can be obtained from the sensing data.

According to the present invention, a respiration measurement system capable of measuring respiration frequency with high accuracy even with sensing data from a wristband type sensor node, and a REM sleep determination system for determining REM sleep (REM sleep) based on the respiration frequency. Can be provided.

1 is an overall view illustrating an example of a configuration of a respiration frequency estimation system according to a first embodiment. It is a figure which shows an example of the bracelet type | mold sensor node 1 which comprises the respiration frequency estimation system of Example 1. FIG. FIG. 2 shows a block diagram of an electronic circuit attached to the substrate 10 of the bracelet type sensor node 1. It is a block diagram which shows each component of the respiration frequency estimation system shown in FIG. It is a flowchart figure which shows the whole flow of the data processing performed with the system of Example 1. FIG. 5 is a flowchart illustrating an example of processing performed by a data totalization program 200 of the PC 103. It is a figure which shows the example of the frequency | count of zero crossing. It is a figure which shows the example of the calculation result of an exercise frequency. It is explanatory drawing which shows the format of the total data table. It is a flowchart which shows an example of the process performed with the sleep period extraction program 300 of PC103. It is explanatory drawing which shows the format of the sleep period data table. It is a flowchart which shows an example of the process performed with the respiration estimation program 400 of PC103. It is explanatory drawing which shows the format of the respiration estimation data table. It is a screen image of the sleep display screen 1200 displayed on the display part 1031 of PC103 which is a client computer. It is explanatory drawing of REM sleep determination. It is a block diagram which shows each component of the REM sleep estimation system of Example 2. It is a flowchart figure which shows the whole flow of the data processing performed with the system of Example 2. FIG. 5 is a flowchart illustrating an example of processing performed by a REM estimation program 500 of the PC 104. It is explanatory drawing which shows the format of the REM data table. It is a screen image of the REM display screen 1300 displayed on the display part 1041 of PC104 which is a client computer. It is an example of the frequency spectrum of the scalar value of the acceleration waveform in sleep data. It is explanatory drawing which shows the position of a bracelet type acceleration sensor. It is explanatory drawing of the relationship between respiration and acceleration. It is a relationship diagram of measured respiratory frequency and estimated respiratory frequency. 3 is a functional block diagram of a client PC 103 in Embodiment 1. FIG. 6 is a functional block diagram of a client PC 103 in Embodiment 2. FIG. 5 is a detailed block diagram of a REM sleep determination unit 2604. FIG.

Hereinafter, embodiments of the present invention will be described with reference to the drawings.

As a first embodiment of the present invention, an acceleration data is measured by a sensor attached to a person (user), a respiration frequency during sleep is calculated, and a respiration frequency estimation system (respiration measurement system, life) is presented to the user. Visualization system).
FIG. 1 is an overall view showing an example of the configuration of the respiration frequency estimation system of the present embodiment. The respiration frequency estimation system of the present embodiment uses the bracelet type sensor node 1 having an acceleration sensor as a sensor for detecting the operation (or state) of the user of the system and detects the acceleration of the arm as biological information. Indicates. The bracelet type sensor node 1 is mounted on a user's (or participant's) arm to detect acceleration, and wirelessly transmits the acceleration data detected at a predetermined cycle to the base station 102 via the antenna 101 as sensing data. , To the client computer (PC) 103. If wired communication is possible, the data is transmitted directly to the PC 103 via a USB connection or the like. Regardless of the method of sending the sensed data sequentially regardless of wireless or wired, a method of sending the data accumulated in the sensor node at regular intervals may be used.
In FIG. 1, the PC 103 communicates with a plurality of bracelet type sensor nodes 1, receives sensing data corresponding to the movement of the user from each bracelet type sensor node 1, analyzes the received sensing data, and outputs display data To do. The output display data can be browsed by the client computer (PC) 103 operated by the user.

FIG. 2 is a diagram showing an example of the bracelet type sensor node 1 constituting the respiration frequency estimation system of the present embodiment, and FIG. 2A is a schematic view seen from the front side of the bracelet type sensor node 1. FIG. 2B is a cross-sectional view of the bracelet type sensor node 1 viewed from the side. This bracelet type sensor node 1 mainly measures the movement of the user.
The bracelet type sensor node 1 includes a case 11 for storing a sensor and a control device, and a band 12 for attaching the case 11 to a human arm.
Inside the case 11 is stored a substrate 10 including a microcomputer (not shown), a sensor 6 and the like (not shown) as shown in FIG. As an example of the sensor 6 that measures the movement of the human body (living body), an acceleration sensor that measures the three-axis accelerations XYZ in the drawing is shown. Note that the bracelet type sensor node 1 may include a temperature sensor and a pulse sensor (not shown), measure the user's body temperature and pulse, and output it as sensing data together with the acceleration. Further, the bracelet type sensor node 1 may be provided with a pressure sensor or a capacitance sensor (not shown), and whether or not the user wears the wristwatch type sensor node 1 may be output as a wearing state.

FIG. 3 shows a block diagram of an electronic circuit attached to the substrate 10 of the bracelet type sensor node 1. In FIG. 3, a substrate 10 includes, for example, a wireless communication unit (RF) 2 including an antenna 5 that communicates with a base station 102, a USB communication unit 39 that is wired to a PC 103, a sensor 6 that is an acceleration sensor, A microcomputer 3 that controls the wireless communication unit 2, a real-time clock (RTC) 4 that functions as a timer for intermittently starting the microcomputer 3, a battery 7 that supplies power to each unit, and a power to the sensor 6 A switch 8 for controlling the supply of is provided. Further, a bypass capacitor C1 is connected between the switch 8 and the sensor 6 to eliminate noise and reduce charge / discharge speed to prevent wasteful power consumption. *

For example, the microcomputer 3 interrupts the CPU 34 based on a signal (timer interrupt) from the CPU 34 that executes arithmetic processing, a ROM 33 that stores programs executed by the CPU 34, a RAM 32 that stores data and the like, and an RTC 4. A serial communication interface (SCI) that transmits and receives signals as serial signals between the interrupt control unit 35 to be applied, the A / D converter 31 that converts the analog signal output from the sensor 6 into a digital signal, and the wireless communication unit 2. ) 36, a wireless communication unit 2, a USB communication unit 39, a parallel interface (PIO) 37 that controls the switch 8, and an oscillation unit (OSC) 30 that supplies a clock to each of the units in the microcomputer 3. The above-described units in the microcomputer 3 are connected via a system bus 38. The RTC 4 outputs an interrupt signal (timer interrupt) at a predetermined period set in advance in the interrupt control unit 35 of the microcomputer 3 and outputs a reference clock to the SCI 36. The PIO 37 controls ON / OFF of the switch 8 in accordance with a command from the CPU 34 and controls power supply to the sensor 6.
The bracelet type sensor node 1 is an identifier that activates the microcomputer 3 at a predetermined cycle (for example, 1 second, for example), acquires sensing data from the sensor 6, and identifies the bracelet type sensor node 1 in the acquired sensing data. To the base station 102 with a time stamp.

FIG. 4 is a block diagram showing each component of the respiratory frequency estimation system shown in FIG. Sensing data transmitted from the bracelet type sensor node 1 is accumulated in the sensing data table 1150 of the recording device 1100 of the client computer (PC) 103 via the base station 102. Or you may communicate with PC103 directly via wired communication which is not illustrated.
The PC 103 includes a display device (output device) 1031 that displays various types of information, and an input device 1032 that allows various information to be input by user operations. The display device 1031 may be a printer or an image file output in addition to a display terminal such as a liquid crystal display or a CRT display. The input device 1032 is an input device such as a keyboard and a mouse. The display device 1031 and the input device 1032 may be a single device having both functions, such as a touch panel display.

The PC 103 further includes a processor 107, a memory 108, and a recording device 1100. The recording device 1100 records various programs and various data tables, which will be described later, such as a hard disk drive, a CD-ROM drive, and a flash memory. Various programs and various data tables may be divided and recorded in a plurality of recording devices.

The processor 107 implements various functions by reading various programs recorded in the recording device 1100 into the memory 108 and executing them. Specifically, by executing the data totaling program 200, the sensing data measured by the acceleration sensor of the user's arm is totaled, a total value for each unit time (for example, one minute) is calculated, and the recording device 1100 is stored in the aggregate data table 250. Also, by executing the sleep period extraction program 300, the calculated total value for each unit time is analyzed, and all sleep periods are detected and stored in the sleep period data table 350 of the recording device 1100. Also, by executing the respiration frequency estimation program, the respiration frequency per unit time (for example, one minute) is estimated from the detected sleep period and stored in the respiration frequency data table 450 of the recording device 1100.

In the following, the PC 103 executes the data totaling program 200, the sleep period extraction program 300, and the respiration frequency estimation program 400 at regular intervals, or based on communication with the wristwatch-type sensor node 1, and further inputs An example is shown in which display data is presented on the display device 1031 due to the operation of the device 1032, the activation of the PC 103, or the end of execution of the respiration frequency estimation program 400.

FIG. 5 is a flowchart showing the overall flow of data processing performed in the system of this embodiment. First, an overview of the entire process will be shown with reference to FIG. 5, and a more specific example of each step will be described later.
First, in step S <b> 1, the base station 102 transfers the sensing data transmitted from the bracelet type sensor node 1 to the PC 103 and accumulates the sensing data in the sensing data table 1150 of the PC 103. Further, an identifier assigned to the sensing data for identifying the sensor node and time information indicating the time when the sensing data is acquired are also stored in the sensing data table 1150 corresponding to the sensing data. Furthermore, the PC 103 (for example, the processor 107, the following steps are also the same) executes the data totaling program 200, calculates the exercise frequency per unit time from the sensing data stored in the recording apparatus 1100, and the recording apparatus 1100 Stored in the aggregate data table 250. The data aggregation program 200 may be executed every predetermined period (for example, 5 minutes), may be executed due to the start or end of communication with the bracelet type sensor node 1, or the input device 1032 The operation may be performed as a cause.

Next, in step S <b> 2, the PC 103 executes the sleep period extraction program 300 to detect an area where the user is estimated to be in a sleep state from the aggregate data stored in the aggregate data table 250, and all sleep areas Is stored in the sleep period data table 350 as a set. In addition, the PC 103 detects main sleep that is the maximum sleep period of the day, adds a main sleep tag to the main sleep among the stored sleep areas, and stores it in the sleep analysis data table 350. . The sleep period extraction program 300 may be executed every predetermined cycle (for example, 5 minutes), or may be executed due to the end of the data aggregation program 200.

Next, in step S3, the PC 103 executes the respiration frequency estimation program 400. Here, for each sleep period stored in the sleep data table 350, the PC 103 obtains sensing data within the period from the sensing data table 1150, and a predetermined period (for example, 1 second) (for example, 1 second) , 5 minutes) and a frequency component having the maximum power is detected as a respiratory frequency candidate. For example, if the maximum power value protrudes sufficiently, (60 * frequency) is determined as the minute respiration frequency of the relevant date and time and is stored in the respiration frequency data table 450. A tag is added to indicate that no respiration has been detected, and the respiration frequency data table 450 is stored. The respiration frequency estimation program 400 may be executed every predetermined cycle (for example, 5 minutes), or may be executed due to the end of the sleep period extraction program 300.

Next, in step S4, the respiratory frequency obtained and stored by the respiratory frequency estimation program 400 is presented on the display unit 1031 of the client computer (PC) 103.
FIG. 6 is a flowchart illustrating an example of processing performed by the data totalization program 200 of the PC 103. Each step is executed by the processor 107 of the PC 103.
First, in step S11, sensing data corresponding to a sensor identifier held by the user is read from the sensing data table 1150. Here, the identifier of the sensor owned by the user may be acquired from, for example, the wristwatch type sensor node 1 communicating with the base station 102, or may be an identifier designated by the user with the input device 102, It may be an arbitrary identifier selected from the user sensor correspondence table. The amount of sensing data to be read here is a predetermined period (for example, 5 minutes) that is an aggregation period of sensing data, or everything after the last aggregation time already stored by the execution of the past data aggregation program 200, etc. You only have to set it.

Next, in steps S12 to S14, a total value is calculated for each predetermined time interval (for example, 1 minute) for the acceleration data of the read sensing data. In the present embodiment, the number of zero crosses indicating the frequency of exercise of the wearer (user) of the bracelet type sensor node 1 within a predetermined time interval is used as the total value.
Since the sensing data detected by the bracelet type sensor node 1 includes X, Y, and Z-axis acceleration data, the scalar amount of the X, Y, and Z-axis acceleration = √ (X ^ 2 + Y ^ 2 + Z ^ 2) is calculated (step S12), and the obtained scalar quantity is filtered (bandpass filter) to extract only a predetermined frequency band (for example, 0.1 Hz to 5 Hz) and remove noise components (step S13). ). The data to which the band pass filter is applied may be appropriately stored for later display, for example. Then, as shown in FIG. 7, a value at which the obtained scalar amount passes a predetermined threshold (for example, 0.05 G) is calculated as the number of zero crosses, and the frequency at which the number of zero crosses appears within a predetermined time interval is calculated. This appearance frequency is output as an exercise frequency at a predetermined time interval (1 minute) (step S14). The result of calculating the exercise frequency is data obtained by sorting the exercise frequency for each unit time in time series as shown in FIG. The motion frequency may be other methods, such as counting the number of times that the acceleration value in each direction of X, Y, and Z vibrates positively and negatively (frequency) within a predetermined time in each direction. In this embodiment, since the calculation can be simplified, a method of calculating the number of zero crossings is employed.

Further, a flag representing the state of data within a predetermined time interval is calculated (step S15). First, of the data existing within a predetermined time interval, the ratio of valid data (that is, data in which three-axis data of X, Y, and Z within a predetermined range exist as values other than missing values) is predetermined. If it is less than a threshold (for example, 0.8), the time interval is determined as missing data. When the wristwatch type sensor node 1 outputs the wearing state, when the time interval is not missing and the ratio of wearing data within the predetermined time interval is less than a predetermined threshold (for example, 0.8), The time interval is determined as non-wearing data. As described above, the flag of the data state at the time interval is determined as one of “data present”, “missing value”, and “non-attached”.
By executing the data totaling program 200, the exercise frequency and the data flag are obtained for each predetermined time interval, and total data for each predetermined time interval is generated as shown in FIG. It accumulates together with the identifier 251 of the wearer of the bracelet type sensor node 1 and the identifier 252 of the bracelet type sensor node 1 (step S16).

FIG. 9 is an explanatory diagram showing the format of the total data table 250. A user ID 251 for storing an identifier of a wearer of the bracelet type sensor node 1 (a user of the respiratory frequency estimation system), a sensor data ID 252 for storing an identifier of the bracelet type sensor node 1 included in the sensing data, and a predetermined time interval Stores the measurement date and time 253 for storing the start time (measurement date and time), the exercise frequency 254 for storing the exercise frequency calculated by the execution of the data totaling program 200, and the status flag of the data obtained by the execution of the data totaling program 200 One entry is formed from the flag 255. The user identifier may be referred to from a table (not shown) set in advance based on the identifier of the bracelet type sensor node 1.

FIG. 10 is a flowchart illustrating an example of processing performed by the sleep period extraction program 300 of the PC 103. Each step is executed by the processor 107 of the PC 103.
First, the unit time aggregate data aggregated by the execution of the data aggregation program 200 is read from the aggregate data table 250 (step S21). The amount of the total data read here may be set to all after the end time of the last sleep period already stored by the execution of the past sleep analysis program 300, for example. For example, the aggregate data to be processed may be deleted from the aggregate data table 250, or a processed flag may be added.

Next, in step S22, a period group estimated to be in a sleep state is detected from the read aggregated data. Although the frequency of exercise during sleep is extremely low, the human body does exercise such as turning over during sleep, so the frequency of exercise does not become zero. Several methods for determining sleep are known. For example, the Cole method (Non-patent Document 2) may be applied. The start time and end time of each period detected by such a method are held as a sleep period candidate group in a temporary storage (not shown) or the like. In addition, when the user does not wear a sensor node, the exercise frequency is close to zero and may be determined as sleep. For example, when a wristwatch-type sensor node has a non-wearing determination unit, sleep is not possible. It is also possible to prevent this by treating the data indicating that the flag 258 is not attached when reading the unit time totalization data table 250 as equivalent to data with high exercise frequency.

Next, in step S23, the adjacent sleep period candidates are combined. Depending on the method of sleep detection, for example, even if the user wakes up temporarily to stop the alarm clock and enters sleep again, the sleep period candidates are divided at the time of the temporary wakeup. However, although sleep as a physiological activity is certainly delimited there, there are cases where it is not desirable to demarcate there as a sleep as a living activity. Therefore, when the next sleep area candidate starts within a predetermined time (for example, 30 minutes) after the end of the sleep period candidate, the two sleep period candidates are combined and handled as one large sleep period. Thus, what can be combined in the sleep period candidate group is searched and combined.

Next, in step S24, those that are incompatible as sleep period candidates are excluded. In the method given as an example, first, sleep period candidates whose duration is a predetermined time (for example, 10 minutes) or less are excluded. If the end time of the last sleep period candidate in the candidate group is within a predetermined time (for example, 30 minutes) from the latest measurement time of the total data read by execution of the sleep analysis program 300, the next sleep analysis program Since there is a possibility that it can be combined with the sleep period newly listed as a candidate in the execution of 300, this is also excluded (returned to the next processing). The sleep period candidate group processed as described above is determined as the sleep period group.

Next, in step S25, main sleep is extracted from the sleep period group determined in step S24, the sleep type is determined as "main sleep", and the other sleep periods are determined as "nap". More specifically, first, the calendar date to which each sleep period belongs is calculated. This means that if the end time of the sleep region is a predetermined time, for example, from 0 o'clock to 20 o'clock, it belongs to the same day, and if it is from 20 o'clock to 24 o'clock, it belongs to the next day. This standard is because it can be considered that sleep that ends before 20 o'clock in the ordinary life is included in the nap. And, for example, the sleep region that starts at 17:00 on July 23 and ends at 19:30 belongs to July 23, for example, the sleep region that starts at 16:30 on July 23 and ends at 20:30 It belongs to July 24th. From the earliest calendar date calculated in this way to the latest calendar date, the longest belonging sleep area in each calendar day is derived, and these are determined as “main sleep” on that calendar day. The type of sleep other than the “main sleep” calculated above is determined as “snapping”. Finally, in step S26, the confirmed sleep period group is accumulated in the sleep analysis data table 350 of the recording device 1100 as shown in FIG. At this time, a sleep ID which is an identifier unique within the sleep analysis data table 350 is assigned to each sleep period. This may be selected, for example, by using a value obtained by adding 1 to the sleep ID assigned last.

FIG. 11 is an explanatory diagram showing the format of the sleep period data table 350. User ID 351 for storing the identifier of the wearer of the bracelet type sensor node 1, sleep ID 352 for storing the sleep identifier, sleep start date and time 353 for storing the start time of the sleep period, and sleep for storing the end time of the sleep period One entry is configured from the end time 354 and the sleep type 355 that stores the sleep type (whether it is main sleep or nap).

FIG. 12 is a flowchart illustrating an example of processing performed by the respiration estimation program 400 of the PC 103. Each step is executed by the processor 107 of the PC 103.
In step S31, the sleep period extracted by the sleep period extraction program 300 is read from the sleep period data table 350. The amount of the sleep period read here may be set to, for example, all after the date and time of the last respiratory data already stored by executing the past respiratory estimation program 400.

Next, in step S32 to step S37, each sleep period acquired here is individually processed.
In step S32, sensor data corresponding to the identifier of the wristwatch sensor node 1 worn by the user included in the sleep period acquired in step S31 is received from the sensing data table 1150. Read.
Next, in step S33, the sensor data acquired in step S32 is cut out for every predetermined time (for example, 1 minute) and surrounding data for a predetermined period (for example, 5 minutes) and is converted into a scalar. The scalar amount of the acceleration data of the X, Y, and Z axes of the sensing data detected by the bracelet type sensor node 1 = √ (X ^ 2 + Y ^ 2 + Z ^ 2) is calculated.
Next, in step S34, the obtained scalar quantity is filtered (bandpass filter) to extract only a predetermined frequency band (for example, 0.01 Hz to 1 Hz) and remove noise components.
Next, in step S35, a frequency spectrum is obtained for the scalar quantity filtered in the previous step S34. By performing FFT (Fast Fourier Transform) on the scalar quantity, the intensity at each frequency is calculated as illustrated in FIG. 21a. At this time, the spectrum may be smoothed by averaging the intensity of each frequency including the intensity of the front and rear frequency components.
Next, in step S36, the frequency having the maximum intensity is acquired as the main frequency from the frequency spectrum obtained in the previous step S35.
Next, in step S37, the validity of the main frequency obtained in the previous step S36 as a respiratory frequency is verified.

For example, it is verified whether the main frequency is within a predetermined frequency range (for example, 0.016 Hz to 0.33 Hz). Otherwise, it may be excluded as being too early or too late for respiration. .
As another example, when the intensity of the main frequency is less than or equal to a predetermined threshold value, the possibility of noise may be high and may be excluded.
As another example, the degree to which the intensity of the main frequency (peak) protrudes from other frequencies is evaluated. If the degree of protrusion is small, the possibility of noise is high and may be excluded. As a means for evaluating the degree of protrusion, for example, the following peak score (Formula 2) may be used.

If this peak score falls below a predetermined threshold (for example, 8.0), it is evaluated that the degree of protrusion is insufficient and is excluded.
When excluded under the conditions exemplified above, it is determined that the breathing at the date and time is “not detected”, and when it is not excluded, “detected successfully”.

Next, in step S38, the respiratory frequency per minute is calculated by multiplying the main frequency calculated in the previous step S36 by 60.
Finally, in step S39, the respiration frequency for each minute at each date and time within each sleep period calculated as described above is accumulated in the respiration estimation data table 450 of the recording device 1100. At this time, if it is determined in step S37 that the breathing at the date and time is “undetected”, the respiration frequency holds a value indicating non-detection (for example, “null”), and otherwise, step S38. Holds the respiration frequency for each minute calculated in.

In the above description, the example of detecting the main frequency component using FFT (Fast Fourier Transform) has been shown. However, any method that can detect the intensity of the frequency component included in the acceleration data may be used. Autocorrelation may be obtained. The main frequency component in the case of using autocorrelation may be, for example, the peak of a peak that appears next to the peak after τ = 0. Further, the validity as the respiration frequency in step S37 may include, for example, a case where the correlation coefficient at the selected τ is greater than or equal to a threshold value.

In the above description, one main frequency component is detected, its validity is verified, and when it is valid, the example is adopted as the respiratory frequency. For example, a plurality of protruding frequency components are detected. May be. In that case, for example, as described in step S37 above, the validity of each is verified, and if there are a plurality of valid frequency components, the most appropriate frequency component may be selected, for example ( For example, an index indicating validity may be obtained from the conditions exemplified above), or the frequency component closest to the respiratory frequency detected at the date and time immediately before the date and time may be selected. As another means, the average respiratory frequency (breathing trend) for each elapsed time after the start of bedtime is calculated by averaging the respiratory rate calculated for the wearer so far. The elapsed time after bedtime start may be calculated, and the frequency component closest to the respiratory trend may be selected.

In the above description, an example in which sensing data for a predetermined period (for example, 5 minutes) is cut out every unit time (for example, 1 minute), the respiration frequency is detected, and accumulated in the respiration estimation data table 450 is shown. However, for example, at a finer interval (for example, 1 second), the sensing data for a predetermined period (for example, 5 minutes) is cut out, the respiration frequency is detected, and the unit for each predetermined unit time (for example, 1 minute). By averaging all breathing frequencies detected in time, it may be strong against sudden noise.

FIG. 13 is an explanatory diagram showing the format of the respiration estimation data table 450. User ID 451 for storing the identifier of the wearer of the bracelet type sensor node 1; date and time 452 for storing the date and time of a predetermined time interval; if the estimated respiration frequency or reasonable respiration is not detected, it is not detected A respiration frequency 453 for storing a value (for example, null) representing the value is held.

FIG. 25 is a functional block diagram of the client PC 103 according to the first embodiment.
The client PC 103 includes, for example, a main cycle component detection unit 2501, a respiratory information acquisition success determination unit 2502, and a storage unit 2503. Each unit is realized by the processor 107 executing the respiration estimation program 400 as described above.
The main period component detection unit 2501 periodically collects the sensing data of the past predetermined period and detects the main period component. This corresponds to the processing in steps S32 to S36 in FIG. The respiration information acquisition success determination unit 2502 determines whether or not the main cycle component is effective as a respiration-related periodic component according to a predetermined determination condition. This corresponds to the process of step S37 in FIG. When it is determined that the accumulating unit 2503 is effective as a periodic component depending on respiration, the accumulating unit 2503 accumulates the frequency of the main periodic component in association with time information as a respiration frequency. This corresponds to the processing of the respiration frequency data table 450 in FIG. 4 and step S374 in FIG.

FIG. 14 is a screen image of the sleep display screen 1200 displayed on the display unit 1031 of the PC 103 which is a client computer. The display of the sleep display screen 1200 by the PC 103 may be caused by accepting a display request from the user via the input device 1032, or may be caused by the end of execution of the respiration estimation program 400, for example If the sensing data of the bracelet type sensor node 1 can be acquired in real time by wireless means or the like, it may be caused by the sleep period extraction program 300 detecting that the wearer has woken up. Note that a browser may be adopted as an application that runs on the PC 103, or an application that runs alone may directly display the sleep display screen 1200.

The sleep display screen 1200 is an example of a screen that presents the user with information on the main sleep of the day and the detected respiratory frequency. This includes a date control 1201 that displays the date to be displayed and allows selection, a scalar quantity graph 1202 that shows the daily transition of the scalar quantity calculated in step S13 of the data aggregation program 200, and this date. A sleep period graph 1203 indicating a sleep period detected from the period, a non-wear period, a respiration frequency graph 1204 indicating a detected respiration frequency, a respiration trend of the day, and a recent respiration frequency trend of the person A respiration trend graph 1205 and a sleep memo panel 1206 for displaying numerical data and advice on sleep of the day are provided.
The date control 1201 is a control for indicating the date displayed on the screen, and the previous and next days may be selected by pressing the left and right buttons.

The scalar quantity graph 1202 is a graph for displaying the data obtained by converting the sensing data into a scalar and applying the bandpass filter calculated in step S13 of the data totaling program 200. By displaying this, the user can intuitively know the amount of body movement during sleep and the amount of activity other than sleep. If the amount of data to be displayed (for example, 24 hours x 60 minutes x 60 seconds x 20 samples = 1728000 data points) is larger than the screen width (eg 1280 pixels), it is actually included within the width of one pixel in the screen It is necessary to represent multiple data points. At this time, for example, one of the data points included in the width of one pixel may be arbitrarily selected and drawn as a single point, or all the data points included in the width of one pixel may be drawn. The average value may be drawn as a single point, or, as shown in the drawing example of FIG. 14B, the average value of all the data points included in the width of one pixel is a single color of dark color. While drawing as points (12021), for example, the standard deviation of data points higher than the average value and the standard deviation of data points lower than the average value are calculated, and the points lower than the average value from the standard deviation of points higher than the average value The area up to the standard deviation of the data is filled with a light color (12022), so that the user understands the whole data and the behavior of the data when viewed in more detail (that is, whether or not the average value varies greatly) Can be recalled.

The sleep period graph 1203 is an area for displaying the sleep period calculated by the sleep period extraction program 300. All sleep regions included in the date to be displayed may be painted with a specific color as shown, and the main sleep may be painted with another color. Further, the sleep period may be integrated into the scalar quantity graph 1202 and displayed from above, for example, translucently.
The respiration frequency graph 1204 is a graph for displaying the respiration frequency detected during the sleep period of the day. A line graph is drawn for the area for which the effective respiration frequency is calculated, and the area for which the effective respiration frequency is not calculated is, for example, painted in a specific color as shown in the figure, and is a respiration non-detection period. May be shown.

The respiration trend graph 1205 is a graph for comparing and displaying the respiration trend obtained by interpolating the deficit value of the respiration of the day with the past respiration trend. The respiratory trend of the day is, for example, about the respiratory frequency detected from the main sleep of the day, taking the “elapsed time after going to bed” on the x-axis and the “respiration frequency” on the y-axis, for example, approximating a quadratic regression equation, Based on this, the estimated respiratory frequency from the start to the end of the sleep period may be plotted with a solid line, for example. Thereby, the user can know the estimated value of the respiration rate even for the missing region. The past breathing trend is, for example, a quadratic regression equation with “elapsed time after going to bed” on the x-axis and “breathing frequency” on the y-axis for the breathing frequency included in all sleep periods detected so far for the user. Approximating y = (qA * x ² + qB * x + qC), and based on this equation, the past trend respiratory frequency from the start to the end of the sleep period may be plotted with a broken line, for example. In this way, the user can know that the sleep rate of the day is higher than the sleep rate of the day, for example, the level of the respiratory rate of falling asleep, the rate of decrease in the respiratory rate after going to bed, or the level of the respiratory rate before waking up. I can do it. In the above description, the example using the quadratic regression equation for calculating the trend has been shown. However, any method may be used as long as the respiratory frequency can be interpolated, for example, a linear regression equation. Moreover, when calculating the past respiratory trend, all the past data of the user may be used, or only data within a predetermined range (for example, the past three months) may be used. The user may also be included in the data. For example, when profile information such as gender and age is held for the user, it may be calculated including only the data of the user having a similar profile.

The sleep memo panel 1206 is an area for displaying numerical data relating to the sleep period of the day, and messages and advice to the user based on the numerical data. For example, the respiratory frequency immediately after going to bed may be calculated using the sleep trend calculated above and displayed as “sleeping rate at bedtime”. Furthermore, using the past respiratory trend, the bedtime respiratory rate in normal sleep may also be presented as “normal bedtime respiratory rate”. As another example, the square term of the quadratic regression equation y = (qA * x ² + qB * x + qC) calculated above, that is, qA may be displayed as a “sleep index”. qA is an index indicating how sharply the respiratory rate decreases after going to bed, and since it is known that the respiratory rate decreases as the sleep becomes deeper, by presenting qA as a sleeping index, Users can know how good their sleep is. At this time, the quadratic regression equation may be approximated using the respiratory frequency of all sleep periods, or the quadratic regression equation may be approximated using only the respiratory frequency within a predetermined time (for example, one hour) after going to bed. Also good. Further, qA may be normalized and displayed so as to fall within a range that is easy to grasp, such as 0 to 10 for many people. Further, a sleep index in normal sleep may be calculated using the past respiratory trend and presented as a “normal sleep index”. In addition to this, for example, the respiration frequency immediately before getting up may be presented as “respiration rate before waking up”, or the quadratic regression equation is approximated using only the respiration frequency calculated within one hour before waking up. The value of the square term qA of the next regression equation may be presented as a “wake-up index”.

Furthermore, advice for obtaining good sleep may be presented to the user based on respiratory frequency and sensing data before and after sleep. For example, if the respiratory rate at bedtime is higher than normal and the activity index (eg, the sum of absolute values of the scalar amount) immediately before going to bed is greater than or equal to a predetermined value, for example, immediately before going to bed “You seem to be in bed before you fall asleep because you may have exercised. By displaying advice such as “Let's refrain from exercising before going to bed,” you can expect to have the effect of exercising just before going to bed and refraining from trying to sleep while your breathing rate is increasing. On the other hand, if the respiratory rate at bedtime is higher than normal and the activity index just before going to bed is less than the specified value, for example, you may drink alcohol. By displaying advice such as “Let's refrain from drinking alcohol before going to sleep because it makes sleep less shallow”, it can be expected to refrain from drinking alcohol. In addition, when the sleep index was higher than usual (when sleep was good), for example, “I slept better than usual. It looks like you're tired. ”By working hard, you can expect users to look back on their fatigue. On the other hand, if the sleep index is lower than usual, for example, “It seems that sleep was worse than usual. If you are not sure, take a bath and relax. " For example, if the qA term and the qB term of the sleep trend for the day are both close to 0 (for example, the absolute value is below a predetermined threshold), the respiratory rate has hardly decreased to the end. Because this is a sign that you can see when you catch it, 'Did you catch a cold? By sending advice such as “Please take a rest slowly,” you can expect users to notice changes in their physical condition. Note that a threshold value for displaying these advices can be set in advance.

In the above embodiment, an example of using the three-axis acceleration sensor of the bracelet type sensor node 1 attached to the user's arm to measure the activity state during sleep of the user (human body) as the respiratory frequency estimation system. As shown, any sensor that can detect the activity state of the human body non-invasively may be used, for example, an angular velocity sensor attached to the arm, or a biaxial or non-axial acceleration sensor instead of a triaxial acceleration sensor. good.
Moreover, although the system which estimates the respiratory frequency during sleep was illustrated in the said embodiment, it is effective also in estimating the respiratory frequency during awakening rather than during sleep. For example, the frequency component of respiration may be reflected in the acceleration sensor even when arms are crossed during a meeting or breathing up immediately after exercise, but this can also be detected and presented by the same means as above. .

In this embodiment, an example is shown in which an acceleration sensor attached to the arm is used as an example of a sensor in which the state of breathing is reflected in a piecewise manner, but any sensor that reflects the state of breathing in a piecewise manner can be used. In this embodiment, the respiration frequency can be measured effectively. For example, when measuring the movement of an object with an infrared camera mounted on the bedding, this embodiment can be applied because the movement of breathing is not reflected on the video depending on the sleeping posture. In addition, as disclosed in, for example, Patent Document 2, in recent years, an acceleration sensor or a face-to-face sensor has been mounted on a name tag badge worn by an employee in a company. Although the breathing motion is not reflected in the acceleration data from the attached badge, the breathing motion may be reflected depending on the sitting posture. You can estimate well.

In the above embodiment, the system for presenting the respiratory frequency during sleep to the display device 1031 using a graph or the like has been exemplified. However, for example, a setting of a period in which the user wants to wake up is accepted by means not shown, and When a behavior suggesting a sleep that is infrequent is shown, a system that causes a user to wake up comfortably by using a bell or the like (not shown) may be used. The behavior suggesting shallow sleep may be, for example, when the fraction of the estimated respiration frequency and the past trend exceeds a predetermined threshold, or the respiration frequency of the most recent predetermined period (for example, 5 minutes) is increasing. May be shown.
According to the embodiment described above, the respiration frequency can be accurately estimated from the output data of a sensor in which the respiration movement is not reflected in the data in many cases like the acceleration sensor worn on the arm.

As a second embodiment of the present invention, acceleration data is measured by a sensor attached to a person (user), the respiratory frequency during sleep is calculated, the REM sleep period is estimated from the increase or decrease in the respiratory frequency during sleep, The REM sleep estimation system shown to a user is shown.
First, problems to be solved by this embodiment will be described. Human sleep is divided into REM sleep and NON-REM sleep. Of these, REM sleep accounts for approximately 20% of normal sleep and is a period in which memory is thought to be organized. As for the state of the body, it is known that the pulse and respiration are more disturbed and the frequency is faster than that during NON-REM. For this reason, Patent Document 3 discloses a technique for recognizing a period in which the pulse rate is equal to or greater than a predetermined threshold as REM sleep by utilizing this fact. Since heartbeat and respiration are controlled physiologically in the same system, REM sleep can be detected by the present invention even when the respiration frequency is used instead of the heart rate. For example, the measured value of the respiratory frequency shown in FIG. 23A shows that a mountain is generated every 60 to 90 minutes, but it is REM sleep that causes this mountain.

On the other hand, the patent document 3 does not specifically mention how to set the threshold. In actual sleep, as shown in FIG. 23 (a), the basal respiration frequency (respiration rate during NON-REM) continues to decrease over time from bedtime to wake-up, and the respiration rate during REM also changes accordingly. Will drop. Therefore, it is not possible to set a single REM sleep threshold throughout sleep. This is because the respiratory frequency during REM in the second half of sleep is often about the same as that of NON-REM in the first half of sleep.
In addition, since the absolute value of the basal respiration frequency and how it falls itself varies depending on the person and the season, it is necessary to set a different threshold for each person.

Furthermore, if the respiratory frequency during the sleep period can be measured from end to end as shown in FIG. 23 (a), REM can be estimated by detecting “mountains”. Thus, the estimated respiration rate can be obtained only in pieces from the bracelet type acceleration sensor. Therefore, it is not easy to find a place where the respiratory frequency is higher than normal.
The present inventors use the estimated respiration rate of a certain user for the past week or so to estimate the basal respiration frequency over time of the user, and determine the location where the respiration frequency higher than the basal respiration rate is measured by REM. By recognizing it as sleep, it was found that REM sleep can be estimated with high accuracy even from fragmentary estimated respiratory frequencies obtained from bracelet type acceleration sensors.

Details of this and an example of processing in this embodiment will be described with reference to the drawings.
FIG. 15A shows in black the REM sleep period measured from EOG (electrocardiogram) for a certain sleep. It can be seen that it occurs in about 90 minutes.
FIG. 15B shows the respiration frequency estimated using a bracelet type acceleration sensor during the same period. It is not possible to determine where the respiration frequency is higher than normal just by looking at this.
FIG.15 (c) is the figure which estimated the basic respiration frequency for every elapsed time after bedtime using the said user's sleep for the past one week, and plotted with the respiration frequency. The basal respiratory frequency was estimated by approximating the elapsed time after the start of each main sleep (sleeping) and the quadratic regression equation for the two axes of the measured respiratory frequency. It can be seen that the curve of the basal respiration frequency monotonously decreases from about 16 breathing frequency when falling asleep to about 14 breathing frequency when waking up. This also revealed a region where the respiratory frequency was higher than usual in the sleep (black arrow). And it turns out that they also correspond with the REM sleep period shown to Fig.15 (a).
FIG. 15D is a diagram in which the deviation between the respiratory frequency estimated during the sleep period and the basal respiratory frequency curve is first calculated and further interpolated. The divergence is the fraction (ratio) of the estimated respiratory frequency and the basal respiratory frequency, ie

Define as Furthermore, since this data contains many missing values, smoothing and interpolation are performed by cubic spline interpolation. In the drawing, only portions where the breathing divergence is 1.0 or more, that is, the breathing frequency exceeds the basic breathing frequency are drawn.
FIG. 15E shows a period in which the interpolated respiratory deviation continuously exceeds 1.0 as a REM candidate period. At this stage, various feature quantities related to the REM candidate period, such as duration and maximum respiratory divergence, are calculated.
FIG. 15F shows what is determined to be REM by determining whether each REM candidate period is REM by, for example, known machine learning. This shows that the REM sleep measured in FIG. 15A can be estimated with high accuracy.

An example of a system configuration for realizing the REM sleep estimation system is shown below. The same components as those of the respiratory frequency estimation system according to the first embodiment are denoted by the same reference numerals, and the description of the overlapping configuration and operation is omitted.
FIG. 16 is a block diagram showing each component of the REM sleep estimation system of the present embodiment. Sensing data transmitted from the bracelet type sensor node 1 is accumulated in the sensing data table 1150 of the recording device 1100 of the client (PC) 104 via the base station 102. Or you may communicate directly with PC104 via the wire communication which is not illustrated.

The PC 104 includes a display device (output device) 1041 that displays various types of information and an input device 1042 that allows various information to be input by user operations. The display device 1041 may be a display terminal such as a liquid crystal display or a CRT display, or a printer or an image file output. Further, a speaker that generates sound may be provided. The input device 1042 is an input device such as a keyboard and a mouse. Further, the display device 1041 and the input device 1042 may be a single device having both functions, such as a touch panel display.
The PC 104 includes a processor 107, a memory 108, and a recording device 1100. The recording device 1100 records various programs and various data tables, which will be described later, such as a hard disk drive, a CD-ROM drive, and a flash memory. Various programs and various data tables may be divided and recorded in a plurality of recording devices.

The processor 107 implements various functions by reading various programs recorded in the recording device 1100 into the memory 108 and executing them. Specifically, by executing the data totaling program 200, the sensing data measured by the acceleration sensor of the user's arm is totaled, a total value for each unit time (for example, one minute) is calculated, and the recording device 1100 is stored in the aggregate data table 250. Also, by executing the sleep period extraction program 300, the calculated total value for each unit time is analyzed, and all sleep periods are detected and stored in the sleep period data table 350 of the recording device 1100. Also, by executing the respiration frequency estimation program, the respiration frequency per unit time (for example, one minute) is estimated from the detected sleep period and stored in the respiration frequency data table 450 of the recording device 1100. Also, by executing the REM estimation program 500, the REM sleep period is estimated from the estimated respiratory frequency and stored in the REM data table 550.

In the following, the PC 104 executes the data tabulation program 200, the sleep period extraction program 300, the respiration frequency estimation program 400, and the REM estimation program 500 at regular intervals, or due to communication with the wristwatch type sensor node 1. An example is shown in which display data is displayed on the display device 1041 as a result of the operation of the input device 1042 or the activation of the PC 104 or the end of the execution of the respiration frequency estimation program 400.

FIG. 17 is a flowchart showing the overall flow of data processing performed in the system of this embodiment.
Steps S2.1 to 2.3 may be the same as S1 to S3 in the first embodiment.
In step S2.4, the PC 104 executes the REM estimation program 500, acquires the respiratory frequency detected within the period from the respiratory frequency data table 450 for each sleep period stored in the sleep period data table 350, and further The basal respiration frequency per unit time (for example, 1 second) after going to bed is calculated using, for example, respiration frequency data for a predetermined period in the past (for example, one week), and the measured respiration frequency and basal respiration frequency are calculated. The divergence is calculated as a respiratory divergence, and a period in which the respiratory divergence satisfies a predetermined condition (for example, exceeds a predetermined threshold) is detected as a REM sleep period and stored in the REM data table 550. The REM estimation program 500 may be executed every predetermined cycle (for example, 5 minutes), or may be executed due to the end of the respiration frequency estimation program 400.

Next, in step S2.5, the REM period stored in the REM estimation program 500 is presented on the display unit 1041 of the client computer 104.
In addition, the data totaling program 200, the sleep period extraction program 300, and the respiration frequency estimation program 400 may be the same as the configuration in the first embodiment, for example.

FIG. 18 is a flowchart illustrating an example of processing performed by the REM estimation program 500 of the PC 104.
First, sleep period data (for example, the sleep start time and sleep end time of a predetermined user) extracted by executing the sleep period extraction program 300 is read from the sleep period data table 350 (step S41). The amount of total data read here may be set to, for example, all after the end time of the last REM period already stored by the execution of the past REM estimation program 500.

Next, in steps S42 to S47, each sleep period read here is individually processed.
In step S42, all the respiratory frequencies calculated by the respiratory frequency estimation program 400 included in one sleep period are read.
Next, in step S43, the basal respiration frequency for every elapsed time after going to bed is calculated from the respiration frequency of all main sleeps for a predetermined period (for example, one week) before the sleep period. The basal respiration frequency approximates a quadratic regression equation y = (qA * x ² + qB * x + qC) with the “elapsed time after going to bed” of the respiration frequency for the predetermined period on the x axis and the “respiration frequency” on the y axis. Based on this formula, the basic respiratory frequency from the start to the end of the sleep period is calculated. As the elapsed time after going to bed, the elapsed time from the sleep start time (sleeping time) in one day can be used. The calculation method of the basal respiration frequency is not particularly limited as long as the respiration frequency can be interpolated. For example, a linear regression equation, a cubic spline method, or a moving average may be used. Moreover, when calculating the basal respiration frequency, all the past data of the user may be used, or only data within a predetermined range (for example, one week) may be used. May be included in the data. For example, when profile information such as gender and age is held for the user, the calculation may be performed including only data of users having similar profiles.

Next, in step S44, the difference between the respiratory frequency detected during the sleep period and the basic respiratory frequency calculated in the step is calculated, and the missing value is interpolated. The deviation may be defined by, for example, a fraction, that is, the following expression.

Further, the calculated part of the breathing divergence, that is, the part where no breathing is detected may be interpolated by a method such as a cubic spline method. Also, at this time, if the last detected breath or the next breath is detected before or after a predetermined period (for example, 5 minutes) at a certain missing date and time, the missing value is not interpolated. By leaving it as it is, it is possible to prevent excessive interpolation in an area where the amount of information is small. In addition to missing values, breathing divergence can be smoothed by substituting interpolated values for portions where the respiratory frequency is normally detected. In addition, interpolation may be performed once and then again.

Next, in step S45, a period in which the calculated respiratory divergence continuously exceeds, for example, a predetermined threshold or more (for example, 1.0) is extracted as one REM candidate period.
Next, in step S46, it is determined whether or not all the extracted REM candidate periods are REM, and the REM period is determined. First, various feature amounts related to the period are calculated from each REM candidate period.
For example, the duration of the REM candidate period may be included as the feature amount. Since it is rare for REM sleep to continue for 15 minutes or more in normal sleep, it can be determined that the possibility that the REM candidate period that continues further is actual REM sleep is low.

As another feature amount, for example, the maximum value of respiratory divergence during the REM candidate period may be included. If the maximum value of respiratory divergence is too low (lower than a predetermined threshold), it can be determined that the possibility of REM sleep is low.
Another feature amount may include, for example, an elapsed time until the maximum value of respiratory divergence occurs after the start of the REM candidate period, or a value obtained by dividing this by the duration of the REM candidate period. Respiratory divergence during REM sleep is about the same as the rising period and the falling period. Therefore, if the elapsed time is extremely short immediately after the start or just before the end, it can be judged that the possibility of REM sleep is low. .
As another feature amount, the ratio of the data points interpolated in step S44 during the REM candidate period may be included as the feature amount. If there are many interpolated data points, the reliability of the REM candidate period should be considered low.
As another feature amount, the respiratory divergence during the REM candidate period may include each coefficient of the quadratic regression equation y = (qA * x ² + qB * x + qC), qA, qB, qC as the feature amount. Since the breathing divergence during REM sleep is often a beautiful mountain shape, it is possible to evaluate the mountain shape by using these coefficients. Note that the determination based on the above-described feature amount can be performed based on a predetermined threshold value or the like.
The feature quantity group mentioned in the above description is an example, and other feature quantities can be used.

It is determined whether or not the REM candidate period is REM sleep using a feature vector including one or more of these feature quantities. For the determination, a model learned in advance using REM sleep learning data may be used. As a learning algorithm at this time, for example, a support vector machine (SVM) disclosed in Non-Patent Document 3 may be used, or any algorithm capable of discriminant analysis may be used. The REM candidate period determined as REM sleep in this step is determined as REM sleep.

The determination of REM sleep can also be performed as follows.
For example, when the difference between the actually measured respiration frequency and the basal respiration frequency is greater than a predetermined slope and increases for a predetermined time or more, it is determined that REM sleep is started, and REM sleep is ended based on a predetermined end condition. Here, the termination condition of REM sleep can include a predetermined time elapses from the start of REM sleep. Also, the REM sleep termination condition may include a monotonically decreasing fraction of the measured or interpolated respiratory frequency and the basal respiratory frequency that is less than or equal to a predetermined slope and longer than or equal to a predetermined time. Furthermore, the REM sleep termination condition can include detecting body movements other than respiration based on the sensing data.
Next, in step S47, the REM candidate period determined as REM sleep in the above step is accumulated in the REM data table 550.

In the above description, an example in which REM sleep is extracted by detecting a location where the respiratory frequency is higher than the basic respiratory frequency is shown. In this method, REM sleep cannot be detected in an area (time zone) where breathing is not detected in the first place. On the other hand, it is known that REM sleep occurs with a period of about 90 minutes for many people. Therefore, it is around an integer multiple of a predetermined time interval (for example, 90 minutes) from the found REM sleep (that is, 90 minutes ago, 90 minutes ago, 180 minutes ago, 180 minutes later, 270 minutes ago,...). Etc.) may be interpolated by determining the region as REM sleep. At this time, instead of 90 minutes, an appropriate time interval (REM sleep interval) may be calculated from all REM sleep periods detected for the wearer, and the calculated time interval may be used. As the calculation method, for example, the time interval from all REM sleeps to the next REM sleep is calculated, and by taking the average value of only the time intervals that were, for example, 135 minutes or less, the wearer's REM An average time interval of sleep may be calculated. As another example, the time interval from every REM sleep to the next REM sleep is calculated, and when the time interval is 135 minutes or more, an integer that is less than 135 minutes by dividing the time interval The wearer is selected by taking the average of the result of dividing by the selected number (for example, if the interval is 330 minutes, then 82.5 minutes is obtained by setting the number to 4). A suitable REM sleep time interval is calculated. Note that it may be less than 135 minutes and closest to 90 minutes (for example, if the interval is 330 minutes, it is 110 minutes with gradual number 3, but 90 minutes with 82.5 minutes with gradual number 4) Close to). This time interval may be calculated only from the sleep period that is the object of calculation, or the average of the time intervals calculated so far for the wearer may be used. The duration of REM sleep determined by correction may be a predetermined time (for example, 15 minutes), or may be the average duration of all REM sleep periods detected so far for the wearer.

FIG. 19 is an explanatory diagram showing the format of the REM data table 550. The user ID 551 stores the identifier of the wearer of the bracelet type sensor node 1, and the REM start stores the start date and time of the REM candidate period determined to be REM sleep. It holds a date and time 552 and a REM end date and time 553 that stores the end date and time of the period. Further, when the discrimination algorithm used for discrimination of the REM candidate period can calculate the likelihood and reliability of discrimination, this information may be stored together with the REM period. If a means for interpolating the REM period is provided, the interpolation flag may be stored together with the REM period.

FIG. 26 is a functional block diagram of the client PC 103 according to the second embodiment.
The client PC 103 includes, for example, the respiration measurement system 2500 according to the first embodiment, the sleep detection unit 2601, the basal respiration frequency calculation unit 2602, the divergence calculation unit 2603, the REM sleep determination unit 2604, and the second accumulation unit 2605. Have. Each unit is realized by the processor 107 executing the programs 200, 300, 400, and 500 as described above.
The sleep detection unit 2601 detects sleep from the sensing data. This corresponds to the processing in step S2.2 in FIG. 17 and step S2 in FIG. The basic respiration frequency calculation unit 2602 is a basic unit for each elapsed time after going to bed, based on the respiration frequency and the time information for a predetermined period accumulated in the accumulation unit 2503 of the respiration measurement system 2500. Calculate respiratory frequency. This corresponds to the processing of steps S42 and S43 in FIG. The divergence calculation unit 2603 calculates the divergence between the calculated basic respiration frequency and the respiration frequency actually measured and accumulated in the accumulation unit 2503 for each elapsed time after going to bed. This corresponds to the processing in step S44 in FIG. The REM sleep determination unit 2604 determines REM sleep from the difference between the basic respiration frequency and the actually measured respiration frequency. This corresponds to the processing in steps S45 and S46 in FIG. The second accumulation unit 2605 accumulates the determined REM sleep time information. This corresponds to the processing in step S47 in FIG.

FIG. 27 is a detailed block diagram of the REM sleep determination unit 2604.
The REM sleep determination unit 2604 includes, for example, a REM candidate period extraction unit 2701, a feature amount calculation unit 2702, and a REM candidate period determination unit 2703.
The REM candidate period extraction unit 2701 extracts, as the REM candidate period, a period in which the difference between the respiratory frequency and the basal respiratory frequency is continuously greater than or equal to a predetermined second threshold value. The feature amount calculation unit 2702 calculates a feature vector including the feature amount in the REM candidate period. REM candidate period discriminating unit 2703 Based on the feature vector, it is determined whether or not the REM candidate period is REM sleep.

FIG. 20 is a screen image of the REM display screen 1300 displayed on the display unit 1041 of the PC 104 which is a client computer. The display of the REM display screen 1300 by the PC 104 may be caused by receiving a display request from a user via the input device 1042, or may be caused by termination of execution of the REM estimation program 500. If the sensing data of the bracelet type sensor node 1 can be acquired in real time by wireless means or the like, it may be caused by the sleep period extraction program 300 detecting that the wearer has woken up. Note that a browser may be adopted as an application that runs on the PC 104, or an application that runs alone may directly display the REM display screen 1300.

The REM display screen 1300 is an example of a screen that presents the main sleep of the day, the detected respiration frequency, and information on the REM sleep to the user. This includes a date control 1301 that displays the date to be displayed and allows selection, a scalar quantity graph 1302 that shows the daily transition of the scalar quantity calculated in step S13 of the data totaling program 200, and from this date. Sleep period graph 1303 indicating the detected sleep period and non-wearing period, respiratory frequency graph 1304 indicating the detected respiratory frequency, REM panel 1305 for detecting REM sleep for the day, and numerical data and advice for the sleep of the day The sleep memo panel 1306 is displayed.
Note that the elements 1301, 1302, 1303, and 1304 may be the same as 1201, 1202, 1203, and 1204 of the sleep screen 1200 in the first embodiment, respectively, and thus detailed description is omitted in the following description.

The REM panel 1305 is an area for displaying the REM sleep period estimated by the REM estimation program 500. As shown in the drawing, the REM period (solid line frame) directly estimated from the respiration frequency may be distinguished from the REM period (broken line frame) interpolated every 90 minutes for the missing portion. Further, for example, information such as the start date and time, end date and time, elapsed time, and estimation reliability of the REM period may be displayed within the frame of the REM period.

The sleep memo panel 1306 is an area for displaying numerical data related to the sleep period of the day, and messages and advice to the user based on the data. For example, the time until the first REM period appears after going to bed on that day may be set as the REM latency, and the average value of the REM latency in the past predetermined time (for example, half a year) may be displayed as the normal REM latency. It is known that the REM latency becomes longer when sleepiness is poor and becomes shorter, for example, in depression. For this reason, displaying the REM latency compared to the normal time is an opportunity for the wearer to save his / her physical condition. Further, the time elapsed since the last REM sleep before getting up may be displayed as, for example, a “clean wake-up index”. In general, it is said that as time passes from REM sleep, sleep becomes deeper, and when waking up with deep sleep, waking is not good. Therefore, by displaying the elapsed time from REM sleep, it becomes a chance to look back on how to wake up. In addition to these, the ratio of REM sleep occupying during sleep, the number of REM sleep periods, the average value of the time interval between REM sleeps, and the like may be presented.

Furthermore, comments and advice regarding sleep may be presented to the user based on respiratory frequency and sensing data before and after sleep. For example, if the REM latency is significantly longer than 90 minutes, sleep deprivation is suspected. By presenting comments such as “Looks like sleep deprivation,” the user is given a chance to review their own lifestyle patterns. May be given.

In the above description, the embodiment in which the respiration frequency is calculated from the sensing data of the bracelet type sensor node 1, the REM sleep period is estimated, and presented on the display unit 1041 has been described. When a buzzer or speaker is provided, REM sleep is terminated by generating a sound due to the termination of REM sleep during the time set by the wearer (for example, 6 to 6:30) It is possible to wake up at the timing that is most easily awakened. In addition to the timing when REM sleep ends, an appropriate timing during REM sleep may be used. In addition, as described above, since memory is organized during REM sleep, studying for the examination is conducted all night, and it is desirable that REM sleep is also passed when sleeping as a short break. Therefore, it is possible to efficiently promote the establishment of short-term memory by generating a sound based on the end of the first REM sleep after going to bed.
According to the embodiment described above, it is possible to accurately estimate REM sleep even from a respiratory frequency including many missing values calculated from an acceleration sensor worn on the arm.

(Modification)
The above-mentioned REM sleep determination may not be obtained by taking into consideration data representing respiration (for example, respiration frequency) from the sensing data, and REM sleep may be determined from the sensing data by the same process as described above. For example, the REM sleep determination method is:
Aggregating sensing data from a sensor that is attached to the arm and acquires sensing data indicating the movement of the arm for the past predetermined period, and detecting a main main period component;
A step of accumulating the frequency of the main period component corresponding to the time information when the main period component is in a predetermined frequency range;
Detecting sleep from the sensing data;
Calculating basic frequency data for each elapsed time after going to bed based on the accumulated frequency of the main period component and time information;
Calculating a deviation between the calculated fundamental frequency data and the accumulated frequency of the main period component;
Determining REM sleep based on the calculated divergence;
Storing the determined time information of REM sleep.

1 bracelet type sensor node 6 sensor 10 substrate 11 case 12 band 103 client computer (PC)
107 Processor 108 Memory 1100 Storage Device 250 Total Data Table 350 Sleep Period Data Table 450 Respiration Frequency Data Table 1150 Sensing Data Table 2500 Respiration Measurement System 2501 Main Period Component Detection Unit 2502 Respiration Information Acquisition Success Determination Unit 2503 Accumulation Unit 2601 Sleep Detection Unit 2602 Basal respiration frequency calculation unit 2603 Deviation calculation unit 2604 REM sleep determination unit 2605 Second accumulation unit 2701 REM candidate period extraction unit 2702 Feature quantity calculation unit 2703 REM candidate period determination unit

Claims (19)

1. A sensor that is attached to the arm and acquires sensing data indicating the movement of the arm;
   A main period component detector that periodically collects sensing data of a predetermined period in the past and detects a main period component that is the main period;
   Respiration information acquisition success determination unit that determines whether the main period component is effective as a period component depending on respiration according to a predetermined determination condition regarding the magnitude or frequency of the main period component;
   A respiration measurement system comprising: an accumulator that accumulates the frequency of the main period component as respiration frequency in association with time information when it is determined to be effective as a period component depending on respiration.
2. The respiratory measurement system according to claim 1, wherein the main period component detection unit uses a maximum peak of a calculation result obtained by performing FFT on a scalar value of sensing data as the main period component.
3. The determination condition in the respiration information acquisition success determination unit includes determining that the respiration information is effective as a periodic component if the power value of the main cycle component is equal to or greater than a predetermined first threshold value. Respiratory measurement system as described.
4. The determination condition in the respiratory information acquisition success determination unit is that the power value of the main period component protrudes by a predetermined amount or more in comparison with the power value of the period component in the first range around the main period component. The respiration measurement system according to claim 1, wherein the respiration measurement system includes determining that it is effective as a periodic component depending on respiration.
5. The determination condition in the respiration information acquisition success determination unit includes determining that the respiration information is effective as a periodic component if the frequency of the main cycle component is within a predetermined second range. Respiratory measurement system as described.
6. 2. The respiratory measurement system according to claim 1, wherein the sensor is an acceleration sensor that acquires acceleration data of a worn arm or an angular velocity sensor that acquires angular velocity data of a worn arm.
7. A respiratory measurement system according to claim 1;
   A sleep detector that detects sleep from the sensing data;
   Basal respiration frequency calculation for calculating a basal respiration frequency for every elapsed time after bedtime based on a respiration frequency for a predetermined period accumulated in the accumulation unit of the respiration measurement system and time information. And
   A divergence calculating unit that calculates a divergence between the calculated basic respiration frequency and a respiration frequency that is actually measured and accumulated in the accumulation unit;
   A REM sleep determination unit that determines REM sleep from the difference between the basal respiration frequency and the actually measured respiration frequency;
   The REM sleep determination system provided with the 2nd storage part which accumulate | stores the time information of the determined REM sleep.
8. The basal respiration frequency calculation unit calculates a basal respiration frequency for each elapsed time after bedtime by calculating an average value for each elapsed time after bedtime from the respiration frequencies for a plurality of sleeps accumulated in the accumulation unit. The REM sleep determination system according to claim 7.
9. The basal respiration frequency calculation unit calculates a basal respiration frequency for each elapsed time after going to bed by approximating a respiration frequency for a plurality of sleeps accumulated in the accumulation unit with a polynomial having a predetermined number of terms. 8. The REM sleep determination system according to 7.
10. The REM sleep determination system according to claim 7, wherein the divergence calculation unit interpolates a period during which the respiratory frequency is not accumulated by a predetermined interpolation method.
11. The REM sleep determination system according to claim 10, wherein the interpolation method in the deviation calculation unit is a cubic spline interpolation method.
12. The REM sleep determination system according to claim 7, wherein the divergence calculated by the divergence calculation unit is expressed as a fraction of the actually measured respiration frequency and the basal respiration frequency.
13. The REM sleep determination unit
    A REM candidate period extraction unit that extracts a period in which the difference between the respiratory frequency and the basal respiratory frequency is continuously equal to or greater than the second threshold as a REM candidate period;
    A feature amount calculation unit that calculates a feature vector including the feature amount in the REM candidate period;
    The REM sleep determination system according to claim 7, further comprising: a REM candidate period determination unit that determines whether the REM candidate period is REM sleep based on the feature vector.
14. The feature vector is
    The duration of the REM candidate period, the maximum difference between the measured or interpolated respiratory frequency and the basic respiratory frequency during the REM candidate period, the elapsed time until the maximum respiratory frequency occurs after the start of the REM candidate period, or the REM of this elapsed time Of the ratio when the duration of the candidate period is a parameter, the ratio of the respiratory rate interpolated during the REM candidate period, and each coefficient when the respiratory frequency during the REM candidate period is approximated by a quadratic regression equation The REM sleep determination system according to claim 13, comprising at least one feature amount.
15. The REM sleep determination unit determines the start of REM sleep when the difference between the actually measured respiration frequency and the basal respiration frequency is greater than a predetermined slope and increases for a predetermined time, and the REM sleep is determined based on a predetermined end condition. The REM sleep determination system according to claim 7, wherein the REM sleep determination system is terminated.
16. The REM sleep determination system according to claim 15, wherein the termination condition of REM sleep includes that a predetermined time elapses from the start of REM sleep.
17. The REM sleep determination system according to claim 15, wherein the termination condition of REM sleep includes that the fraction of the actually measured or interpolated respiration frequency and the basal respiration frequency is monotonously decreased for a predetermined time or less with a predetermined inclination or less.
18. The REM sleep determination system according to claim 15, wherein the end condition of REM sleep includes detecting body movements other than respiration based on sensing data.
19. The REM sleep determination system according to claim 7, wherein a period corresponding to a multiple of the REM sleep interval is also determined as REM sleep according to a REM sleep interval based on a plurality of REM sleep intervals determined by the REM sleep determination unit.

Images