JP2012019927A - Cycle estimating device, cycle estimating method, and program - Google Patents

Abstract

An exercise cycle (walking cycle) that can be applied to acceleration data at a low sampling rate and shows a period corresponding to one set of two movements with a small amount of calculation depends on one of the two. Even if the acceleration data is larger than the other acceleration data and the distortion data is large, it is estimated with high accuracy.
An AMDF calculation unit 15 performs an AMDF calculation on acceleration data acquired by a carried acceleration measurement unit. The period determination unit 17 calculates a period for one step of the walking motion based on the minimum value of the AMDF result calculated by the AMDF calculation unit 15 and the minimum value in the extraction range of the AMDF result determined based on the minimum value. Estimate the walking cycle shown.
[Selection] Figure 1

Description

  The present invention relates to a period estimation device, a period estimation method, and a program for estimating a user's motion state using a mobile terminal device equipped with an acceleration sensor.

  To realize a context-aware service, context estimation by a mobile terminal device equipped with a sensor is becoming widespread. In particular, the use of a system in which an acceleration sensor is mounted on a mobile terminal device and a user's motion state such as walking or running is estimated is rapidly expanding. In such estimation of the motion state, the frequency of motion is one of important information. In particular, in estimating the walking state, not only know how fast you are walking, but also a frequency that accurately reflects the movement of each step, assuming rehabilitation and support for sports. Information is important.

  In a context estimation system using a portable terminal device, estimation calculation is performed by the portable terminal device from the viewpoint of a radio wave reach and communication power consumption. When the estimation calculation is performed by the mobile terminal device, not only can the estimation be performed with high accuracy, but in order to increase the battery duration, it is necessary to reduce the calculation amount and reduce the power consumption. Further, it is difficult to increase the sampling rate of the acceleration sensor up to several hundred Hz because power consumption increases in many devices.

  In the estimation of the frequency of the walking state using an acceleration sensor, the most general method is a method of transforming into the frequency domain using FFT (Fast Fourier Transform). However, FFT requires a large amount of calculation and increases power consumption.

  On the other hand, as a method for obtaining a frequency with a small amount of calculation, there is a method using AMDF (Average Magnitude Difference Function) used in the voice field (for example, see Non-Patent Document 1). The AMDF is a sum of absolute values of differences between a signal on the time axis and a signal obtained by delaying the signal. In this case, the delayed time is called lag. When the lag is an integer multiple of the period of the fundamental frequency of the signal, the AMDF result is minimal. In the estimation of the fundamental frequency by the AMDF, the point having the smallest lag among the points where the AMDF result is the smallest or the value near the smallest is obtained, and the lag at that point is set as the period.

M. Ross, H. Shaffer, A. Cohenand, R. Freudberg, and H. Manley, "Average magnitude difference function pitch extractor", IEEE transactions on acoustics, speech and signal processing, vol. 22, No. 5, pp. 353-362, 1974.

  However, in the method using AMDF, the frequency cannot be accurately obtained unless the data has a high sampling rate. As described above, AMDF is intended for data having a high sampling rate of several hundred times the fundamental frequency as in the audio field. Therefore, when applied to data at a low sampling rate, the lag with the smallest AMDF result does not always correctly represent the period, and it is difficult to obtain the period.

  Therefore, for walking motion (a pair of motion motion), the acceleration data is acquired at a low sampling rate of about 10 times the cycle in walking motion, and AMDF is simply applied, and the AMDF result is minimized. However, there is a problem that it is difficult to obtain an accurate period. In particular, when data distortion is large, there is a problem that there is a high possibility that an error will increase.

  The present invention has been made in consideration of such circumstances, and the object thereof can be applied to acceleration data at a low sampling rate, and a pair of motion motions 1 with a small amount of calculation. A period estimation device that can accurately estimate a motion cycle indicating a period of time even if the acceleration data from one of the two is larger than the other acceleration data, even if the data has a large distortion Another object is to provide a period estimation method and a program.

  In order to solve the above-described problem, the present invention provides a cycle estimation method for estimating a motion cycle indicating a period of one pair of motion motions with respect to acceleration data acquired by a portable acceleration measurement unit. An apparatus that performs an AMDF (Average Magnitude Difference Function) operation on the acceleration data, a minimum value of an AMDF result calculated by the calculation unit, and an AMDF result determined based on the minimum value A period estimation apparatus comprising: a period estimation unit configured to estimate the movement period based on a minimum value in an extraction range.

  Further, the present invention is the above invention, wherein the period estimation device includes a dividing unit that divides the acceleration data into divided data of a predetermined section, and the arithmetic unit converts the divided data divided by the dividing unit into divided data. It is characterized in that it performs AMDF operation on it.

  Further, according to the present invention, in the above invention, the period estimation unit is configured such that a difference from the minimum value of the calculated AMDF result among the minimum values in the calculated AMDF result is equal to or less than a first threshold value set in advance. It is determined whether or not there is a local minimum value, and if there is a local minimum value that is less than or equal to the first threshold value, the minimum among the lags corresponding to the local minimum value that is less than or equal to the first threshold value. If the range below the lag is set as the extraction range, and there is no minimum value below the first threshold, the range below the lag corresponding to the calculated minimum value of the AMDF result is the extraction range. It is characterized by setting to.

  Also, in the present invention according to the invention described above, the period estimation unit extracts the AMDF result in the extraction range, and further, among the local minimum values in the extraction range, there is a difference from the minimum value of the extracted AMDF result. It is determined whether or not there is a minimum value that is less than or equal to a preset second threshold value, and if there is no minimum value that is less than or equal to the second threshold value, a lag that takes the minimum value of the extracted AMDF result Is estimated as the movement period, and if there is a minimum value that is less than or equal to the second threshold, among the minimum values that are less than or equal to the second threshold, half of the lag that takes the minimum value of the AMDF result The minimum value lag closest to the value is estimated as the motion period.

  Moreover, the present invention is characterized in that, in the above invention, the motion cycle is a walking cycle indicating a period of one step of walking motion.

  Further, in order to solve the above-described problem, the present invention estimates a motion cycle indicating a period of one pair of motion motions with respect to acceleration data acquired by a carried acceleration measurement unit. A method for estimating a period, the step of performing an Average Magnitude Difference Function (AMDF) operation on the acceleration data, a minimum value of an AMDF result calculated by the calculation unit, and an AMDF result determined based on the minimum value And a step of estimating the motion period based on the minimum value in the extraction range.

  Further, the present invention is the above invention, wherein the period estimation method includes a step of dividing the acceleration data into divided data of a predetermined section, and the step of performing the AMDF operation is performed on the divided divided data. And performing an AMDF operation.

  Also, in the present invention according to the above invention, the step of estimating the movement period includes a first difference that is set in advance, wherein a difference from a minimum value of the calculated AMDF result is a preset minimum value in the calculated AMDF result. Determining whether or not there is a minimum value that is less than or equal to the threshold value, and if there is a minimum value that is less than or equal to the first threshold value, the lag corresponding to the minimum value that is less than or equal to the first threshold value And setting the range below the minimum lag as the extraction range, and if there is no minimum value below the first threshold, the lag corresponding to the minimum value of the calculated AMDF result And a step of setting the following range as the extraction range.

  Further, the present invention is the above invention, wherein the step of estimating the motion period includes a step of extracting an AMDF result in the extraction range, and a minimum value of the extracted AMDF result among the minimum values in the extraction range. Determining whether there is a local minimum value that is less than or equal to a preset second threshold value, and if there is no local minimum value that is less than or equal to the second threshold value, The step of estimating the lag taking the minimum value as the motion period and the minimum value of the AMDF result among the minimum values of the second threshold value or less when there is a minimum value of the second threshold value or less. Estimating a lag having a minimum value closest to a half value of the lag taking a value as the motion period.

  Moreover, the present invention is characterized in that, in the above invention, the motion cycle is a walking cycle indicating a period of one step of walking motion.

  Further, in order to solve the above-described problem, the present invention estimates a motion cycle indicating a period of one pair of motion motions with respect to acceleration data acquired by a carried acceleration measurement unit. In a computer as a cycle estimation device, a calculation unit performs an AMDF (Average Magnitude Difference Function) operation on the acceleration data, and a cycle estimation unit includes a minimum value of the AMDF result calculated by the calculation unit, and A program for executing a period estimation procedure for estimating the motion period based on the minimum value in the extraction range of the AMDF result determined based on the minimum value.

  According to this invention, a calculating part performs AMDF (Average Magnitude Difference Function) calculation with respect to the acceleration data acquired by the carried acceleration measuring part. Based on the minimum value of the AMDF result calculated by the calculation unit and the minimum value in the extraction range of the AMDF result determined based on the minimum value, the cycle estimation unit is equivalent to one pair of exercise motions. Estimate the period of motion indicating the period. For this reason, it can be applied to acceleration data at a low sampling rate, and with a small amount of calculation, the acceleration cycle by one of the two acceleration data by one of the two sets of exercise motion is the other. Even high-distortion data, which has been observed larger than the acceleration data, can be estimated with high accuracy.

It is a block diagram which shows the structure of the portable terminal device 1 by embodiment of this invention. It is a flowchart for demonstrating operation | movement (period determination process) of the portable terminal device 1 by this embodiment. When performing period determination using AMDF for data that is not low frequency, as an example that requires a high sampling rate, an example in which AMDF is performed on audio data that is sampled at 11025 Hz for audio having a fundamental frequency of 87 Hz FIG. When performing period determination using AMDF for data that is not low frequency, as an example in which a high sampling rate is required, an example in which AMDF is performed on audio data obtained by sampling a voice with a basic frequency of 87 Hz at 1002 Hz. FIG. It is a conceptual diagram which shows acceleration data when the portable terminal device 1 containing the acceleration measurement part (acceleration sensor) 10 walks in a breast pocket, and is sampled at 19.6 Hz. In this embodiment, it is a conceptual diagram which shows the window division | segmentation data obtained by dividing | segmenting the said acceleration data into a window. In this embodiment, it is a conceptual diagram which shows the result of having performed AMDF with respect to the said window division | segmentation data. In the present embodiment, a result obtained by extracting only the AMDF result having the minimum lag or less among the lags corresponding to the minimum value having a difference from the minimum value equal to or less than the constant k with respect to the extracted AMDF result is shown. It is a conceptual diagram. In this embodiment, it is a conceptual diagram which shows the acceleration data when the portable terminal device 1 containing the acceleration measurement part (acceleration sensor) 10 walks in the pocket of a pants, and samples at 19.6 Hz. In this embodiment, it is a conceptual diagram which shows the window division | segmentation data obtained by dividing | segmenting the said acceleration data into a window. In this embodiment, it is a conceptual diagram which shows the result of having performed AMDF with respect to the said window division | segmentation data. In this embodiment, it is a conceptual diagram which shows the result of having extracted the said extracted AMDF result. In this embodiment, the average estimation error when measuring the acceleration at the time of walking and running at a distance of 100 m for 16 subjects and estimating the gait using each of the FFT and the AMDF according to the present embodiment. FIG.

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

  FIG. 1 is a block diagram showing a configuration of a mobile terminal device 1 according to an embodiment of the present invention. In FIG. 1, a mobile terminal device 1 (period estimation device) includes an acceleration measurement unit 10, a sampling unit 11, a sensor data storage unit 12, a window division unit 13, a window division data storage unit 14, and an AMDF (Average Magnitude Difference Function) calculation. Unit 15, AMDF storage unit 16, cycle determination unit 17, cycle storage unit 18, data transmission unit 19, and power supply unit 20.

The acceleration measuring unit 10 includes a so-called acceleration sensor, measures the acceleration of the mobile terminal device 1, and outputs the acceleration to the sampling unit 11.
The sampling unit 11 samples the output (acceleration) of the acceleration measuring unit 10 at a predetermined sampling rate. For example, the sampling unit 11 performs sampling at about 20 Hz, which is a sampling rate of about 10 times the period of 0.5 s (frequency: 2 Hz (hertz)) of one walking motion. The sampling unit 11 stores the sampled acceleration data in the sensor data storage unit 12. Further, the sampling unit 11 reads the acceleration data stored in the sensor data storage unit 12 and supplies the acceleration data to the window dividing unit 13.
The sensor data storage unit 12 is connected to the sampling unit 11 and stores sampled acceleration data.

The window dividing unit 13 (dividing unit) divides temporally continuous data into windows (small sections) for acceleration data with a low sampling rate of about 20 Hz supplied from the sampling unit 11. That is, the window dividing unit 13 divides the acceleration data into divided data of a predetermined section. Further, the window division unit 13 stores the window division data of the divided windows (small sections) in the window division data storage unit 14. The window division unit 13 reads the window division data stored in the window division data storage unit 14 and supplies the window division data to the AMDF calculation unit 15.
The window division data storage unit 14 is connected to the window division unit 13 and stores window division data divided into windows.

The AMDF calculation unit 15 (calculation unit) performs an AMDF calculation on the window division data supplied from the window division unit 13. That is, the AMDF calculation unit 15 performs the AMDF calculation on the divided data divided by the window division unit 13. The AMDF calculation unit 15 stores the calculated AMDF calculation result in the AMDF storage unit 16. Further, the AMDF calculation unit 15 reads out the AMDF calculation result stored in the AMDF storage unit 16 and supplies the AMDF calculation result to the period determination unit 17.
The AMDF storage unit 16 is connected to the AMDF calculation unit 15 and stores the AMDF calculation result.

The period determination unit 17 (period estimation unit) performs the walking motion 1 based on the minimum value of the AMDF result calculated by the AMDF calculation unit 15 and the minimum value in the extraction range of the AMDF result determined based on the minimum value. A walking cycle that is a cycle of steps (an exercise cycle indicating a period of one pair of exercise operations) is estimated. That is, the walking cycle indicates a period of one step of walking motion.
Further, the period determination unit 17 has a minimum value in which the difference from the minimum value of the calculated AMDF result is equal to or less than a preset constant k (first threshold) among the minimum values in the calculated AMDF result. It is determined whether or not. When there is a local minimum value that is a constant k (first threshold value), the cycle determination unit 17 determines that the minimum is the lag corresponding to the local minimum value that is equal to or less than the constant k (first threshold value). The range below the lag is set as the above-described extraction range. In addition, when there is no minimum value that is equal to or less than the constant k (first threshold), the period determination unit 17 sets a range that is equal to or less than the lag corresponding to the calculated minimum value of the AMDF result as the above-described extraction range. To do.

Further, the period determining unit 17 extracts the AMDF result in the extraction range described above, and further, among the local minimum values in the extraction range, the difference from the minimum value of the extracted AMDF result is a preset constant c (second threshold value). ) It is determined whether or not there is another local minimum value as follows. When there is no other minimum value, the period determining unit 17 estimates a lag that takes the minimum value of the extracted AMDF result as a period for one step of the walking motion. In addition, when other local minimum values exist, the cycle determination unit 17 uses the local minimum lag closest to half the value of the lag that takes the minimum value of the AMDF result among the other local minimum values as the walking motion 1. Estimated as the cycle of the steps (movement cycle). Note that the minimum value of the AMDF result is also a minimum value. Here, the other minimum value is another minimum value different from the minimum value of the AMDF result.
Note that the cycle determination unit 17 causes the cycle storage unit 18 to store the estimated cycle (movement cycle) for one step of walking. In addition, the cycle determination unit 17 reads the data from the cycle storage unit 18 as a cycle for one step of walking (exercise cycle), and supplies the cycle of one step of the walking operation (exercise cycle) to the data transmission unit 19.

  That is, the period determination unit 17 obtains the minimum value of the AMDF result, and among the lag corresponding to this minimum value and the lag corresponding to another local minimum value whose difference from the minimum value is a constant k or less, Only the AMDF results below the lag are extracted. The period determination unit 17 further determines whether or not there is another minimum value whose difference from the extracted minimum value of the AMDF result is a constant c or less. When there is no cycle, the cycle determination unit 17 sets a lag that takes the minimum value of the extracted AMDF result as a cycle (motion cycle) for one step of walking motion. If present, the period determining unit 17 regards the period as two steps of walking movement (period of two movement movements), and minimizes the difference between the extracted minimum value of the AMDF result and the constant c or less. Among these, the minimum value lag closest to half the value of the lag taking the minimum value of the AMDF result is set as a cycle (motion cycle) for one step of the walking motion.

The cycle storage unit 18 is connected to the cycle determination unit 17 and stores the lag obtained by the cycle determination unit 17 as a cycle (motion cycle) for one step of walking motion.
The data transmission unit 19 reads the cycle (exercise cycle) for one step of walking motion stored in the cycle storage unit 18 via the cycle determination unit 17 and outputs the cycle (exercise cycle) for one step of walking motion. .
The power supply unit 20 supplies power to each unit in the apparatus.

  FIG. 2 is a flowchart for explaining the operation (period determination process) of the mobile terminal device 1 according to the present embodiment. In the period determination process, first, the sampling unit 11 samples the output of the acceleration measuring unit 10 at a low sampling rate of about 20 Hz and stores it in the sensor data storage unit 12 (step S1). Next, the window dividing unit 13 first divides temporally continuous data into windows (small sections) for the stored acceleration data at a low sampling rate such as 20 Hz, and stores the window divided data as window divided data. It memorize | stores in the part 14 (step S2).

Next, the AMDF operation unit 15 performs an AMDF operation on the stored window division data, and stores the obtained AMDF result in the AMDF storage unit 16 (step S3). Next, the period determination unit 17 obtains the minimum value of the AMDF result in the stored AMDF, and the difference between the lag corresponding to the minimum value and the minimum value and the minimum value is a constant k (first threshold value). Of the lags corresponding to the local minimum values below, only the AMDF results below the minimum lag are extracted as the extraction range (step S4). The constant k is artificially set in advance from the tendency of the collected sensor data, or is determined in advance by a machine learning algorithm or the like.
The flowchart described here is an example in the case where there is a minimum value in which the difference between the minimum value and the minimum value of the AMDF result is equal to or less than a constant k (first threshold). If there is no minimum value in which the difference between the minimum value and the minimum value of the AMDF result is equal to or less than the constant k (first threshold), in step S4, the period determination unit 17 determines the above extraction range as the AMDF result. It is extracted as the range below the lag corresponding to the minimum value of.

  Further, the cycle determination unit 17 determines whether or not there is a minimum value in which the difference from the minimum value is equal to or less than a constant c (second threshold) in the extracted AMDF result (step S5). Similarly to the constant k, the constant c is set in advance from prior knowledge of data. If there is no local minimum value that makes the difference from the minimum value equal to or less than the constant c (NO in step S5), the period determination unit 17 sets the lag for the minimum value of the AMDF result as one step of the walking motion. It memorize | stores in the period memory | storage part 18 as a period (step S6).

  On the other hand, when there is a minimum value where the difference from the minimum value is equal to or less than the constant c (YES in step S5), the acceleration when either the left or right foot moves has been observed to be larger than the other. It is considered that the data has a large distortion. Therefore, the period determination unit 17 regards the lag that takes the minimum value of the AMDF result as the period of two steps of walking motion, and determines the minimum value of the AMDF result from among the minimum values whose difference from the minimum value is a constant c or less. The minimum value lag that is the lag closest to the half value of the lag to be taken is stored in the cycle storage unit 18 as a cycle for one step of the walking motion (step S7).

  In any case, the data transmission unit 19 outputs the cycle stored in the cycle storage unit 18 (step S8).

  FIG. 3 and FIG. 4 show audio data obtained by sampling audio with a fundamental frequency of 87 Hz at 11025 Hz and 1002 Hz as an example in which a high sampling rate is required when performing period determination using data with non-low frequency data. It is a conceptual diagram which shows the example which performed AMDF with respect to. For audio data, a file recorded in monaural WAVE format was used, and AMDF was directly applied to the quantized value at the time of recording. As for the number of peaks and valleys of data used when obtaining an AMDF result, 256 amplitude difference data are used at 11025 Hz and 32 amplitude differences at 1002 Hz so that there is no significant difference depending on the sampling rate. The data was used to determine AMDF results.

  In conclusion, since data of 87 Hz is used, 11.5 ms (milliseconds) needs to be obtained as a cycle. As shown in FIG. 3, when AMDF is performed on audio data obtained by sampling 87 Hz audio at 11025 Hz, a point that is minimized is easily obtained. On the other hand, as shown in FIG. 4, when AMDF is applied to audio data sampled at 1002 Hz for 87 Hz audio, a minimum value near 11.5 ms and a lag less than 11.5 ms are obtained. The local minimum value cannot be distinguished, and the period cannot be obtained correctly. As described above, with voice, it is not possible to determine the correct period for data having a sampling rate of about 10 times the signal to be obtained by the method using AMDF.

  On the other hand, as an example of processing for low-frequency data with a small distortion, which is the subject of the present invention, the portable terminal device 1 including the acceleration measuring unit (acceleration sensor) 10 is walked in a breast pocket and sampled at 19.6 Hz. An example in which data is acquired at a rate and cycle determination is performed will be described. In this case, since the average frequency of one step of walking of an adult male is about 2 Hz (cycle: 0.5 s (seconds)), data is acquired at a sampling rate about 10 times that frequency. Further, the constant k in the period determination is set to 100 mG (milli gie), and the constant c is set to 200 mG.

  FIG. 5 is a conceptual diagram showing acceleration data when the portable terminal device 1 including the acceleration measuring unit (acceleration sensor) 10 is walked in a breast pocket and sampled at 19.6 Hz. FIG. 6 is a conceptual diagram showing window division data obtained by dividing the acceleration data into windows in the present embodiment. In this case, the number of sensor data samples cut out by the window is, for example, 64. The number of samples 64 at the sampling rate of 19.6 Hz corresponds to a time range of about 3.3 s, and a sufficient range is cut out for 0.5 s of the cycle of one step of walking motion.

  FIG. 7 is a conceptual diagram showing a result of performing AMDF on the window division data in the present embodiment. In this case, an AMDF result was obtained using 32 amplitude difference data at each lag of 51 ms to 1600 ms. In this AMDF result, the minimum value is 200 mG, and the lag corresponding to this minimum value 200 mG is 1.0 s. Comparing the minimum value of 200 mG with other minimum values of AMDF, it can be seen that the difference between the minimum value of 0.5 s and 1.45 s and the minimum value is a constant k = 100 mG or less. Therefore, only the AMDF result of 0.5 s or less, which is the smallest lag among 0.5 s, 1.0 s, and 1.45 s, is extracted.

  FIG. 8 shows that, in the present embodiment, only the AMDF results less than the minimum lag among the lags corresponding to the minimum values where the difference from the minimum value is less than or equal to the constant k is extracted from the extracted AMDF results. It is a conceptual diagram which shows the result. In FIG. 8, in the extracted AMDF result, the minimum value is 280 mG, and the lag corresponding to this minimum value 280 mG is 0.5 s. Further, for this extracted AMDF result, a local minimum value whose difference from the minimum value is a constant c = 200 mG or less is searched, but such a local minimum value does not exist. Therefore, 0.5 s taking the minimum value of the AMDF result is set as a cycle of one walking motion.

  On the other hand, as an example of processing for low-frequency data with large distortion, the portable terminal device 1 including the acceleration measuring unit (acceleration sensor) 10 is walked in a pants pocket, data is acquired at a sampling rate of 19.6 Hz, and period determination is performed. An example of performing is described. In this case, the constant k and the constant c in the period determination process are set to 100 mG and 200 mG, respectively, in the same manner as when placed in the breast pocket.

  FIG. 9 is a conceptual diagram showing acceleration data when the portable terminal device 1 including the acceleration measuring unit (acceleration sensor) 10 is walked in a pants pocket and sampled at 19.6 Hz in the present embodiment. FIG. 10 is a conceptual diagram showing window division data obtained by dividing the acceleration data into windows in the present embodiment. In this case, the number of sensor data samples cut out by the window was 64.

  FIG. 11 is a conceptual diagram showing a result of performing AMDF on the window division data in the present embodiment. In this case, an AMDF result was obtained using 32 amplitude difference data at each lag of 51 ms to 1600 ms. In this AMDF result, the minimum value is 300 mG, and the lag corresponding to this minimum value 300 mG is 1.0 s. When the minimum value of 300 mG is compared with the minimum value of other AMDF results, there is no other minimum value whose difference from the minimum value is a constant k = 100 mG or less. Therefore, only an AMDF result of 1.0 s or less, which is a lag corresponding to the minimum value, is extracted.

  FIG. 12 is a conceptual diagram illustrating a result of extracting the extracted AMDF result in the present embodiment. As for the extracted AMDF results shown in FIG. 12, when searching for another minimum value where the difference from the minimum value is a constant c = 200 mG or less, the difference between the minimum value of 0.5 s and 0.9 s is the minimum value. It can be seen that the constant c is 200 mG or less. For this reason, this acceleration data is considered to be data with a large distortion in which the frequency per two steps is detected stronger than the frequency per one step of the walking motion. Therefore, among the local minimum values, 0.5 s, which is the cycle closest to half of 1.0 s, which is the lag taking the minimum value of the AMDF result, is set as the cycle of one step of walking motion.

  FIG. 13 shows a case in which the acceleration at the time of walking and running at a distance of 100 m is measured for 16 subjects in this embodiment, and the gait estimation is performed using each of the FFT and the AMDF according to the present embodiment. It is a table | surface figure which shows the average estimation error of. In this case, acceleration data was acquired at 19.6 Hz, and 64 samples were cut out with a window. On the other hand, FFT was performed at 64 points, and the AMDF according to the present embodiment was performed using 32 data at each lag.

  The portable terminal device 1 including the acceleration measuring unit (acceleration sensor) 10 was carried in three places in a trouser pocket, a breast pocket, or a bag, and experiments were conducted. In FIG. 13, the value after ± is a standard deviation, and represents a variation in estimation accuracy among subjects. Under all conditions, it was found that the error was smaller when using AMDF than when using FFT. In particular, in the case of carrying in a trouser pocket where the waveform distortion is large, the error improvement by the AMDF according to the present embodiment is large.

According to the embodiment described above, the window dividing unit 13 divides acceleration data into divided data of a predetermined section. The AMDF operation unit 15 performs an AMDF operation on the divided data divided by the window division unit 13. Based on the minimum value of the AMDF result calculated by the AMDF calculation unit 15 and the minimum value in the extraction range of the AMDF result determined based on the minimum value, the cycle determination unit 17 determines the cycle for one step of the walking motion. presume. Thus, even when acceleration data having a low sampling rate of about 10 times the period to be estimated is used, it is possible to perform a highly accurate period estimation of one step of walking motion using AMDF. In addition, by using AMDF instead of FFT, the cycle of acceleration data can be obtained with a small amount of calculation. For this reason, power consumption in the mobile terminal device 1 can be reduced.
In addition, the period of one step of walking motion should be estimated with high accuracy even if the acceleration data when one of the left and right feet moves is observed to be larger than the other acceleration data. Can do.
That is, the mobile terminal device 1 can estimate the cycle of one step of walking with high accuracy with a small amount of calculation even if the acceleration data has a low sampling rate.

  Further, unlike the FFT that extracts only the strongest frequency among the harmonics, the AMDF obtains a fundamental frequency corresponding to the distance on the frequency axis between the harmonics, and thus is more resistant to acceleration data having a larger distortion than the FFT.

Note that, according to the embodiment of the present invention, the mobile terminal device 1 has a walking cycle (a set of two) indicating a period of one step of the walking motion with respect to the acceleration data acquired by the carried acceleration measurement unit 10. Is a period estimation device that estimates the period of one motion movement), and is calculated by the AMDF calculation unit 15 (calculation unit) that performs AMDF calculation on the acceleration data, and the AMDF calculation unit 15 A cycle determination unit 17 (cycle estimation unit) that estimates the walking cycle (motion cycle) based on the minimum value of the AMDF result and the minimum value in the extraction range of the AMDF result determined based on the minimum value.
Thereby, even if it is a case where the acceleration data of the low sampling rate of about 10 times with respect to the period to estimate is used, the portable terminal device 1 can estimate the period of one step of walking motion with AMDF (exercise period estimation). )It can be performed. In addition, in the mobile terminal device 1, by using AMDF instead of FFT, it is possible to obtain a cycle (motion cycle) for one step of walking motion with a small amount of calculation. For this reason, power consumption in the mobile terminal device 1 can be reduced.

Further, the mobile terminal device 1 includes a window dividing unit 13 (dividing unit) that divides acceleration data into divided data of a predetermined section, and the AMDF calculating unit 15 applies the divided data divided by the window dividing unit 13 to the divided data. Perform an AMDF operation.
Thereby, the amount of calculation of AMDF can be further reduced.

Further, the period determination unit 17 has a minimum value in which the difference from the minimum value of the calculated AMDF result is equal to or less than a preset constant k (first threshold) among the minimum values in the calculated AMDF result. If there is a minimum value that is less than or equal to the constant k, among the lags corresponding to the minimum value that is less than or equal to the constant k, the range below the minimum lag is set as the extraction range. If there is no minimum value that is less than or equal to k, a range that is less than or equal to the lag corresponding to the calculated minimum value of the AMDF result is set as the extraction range.
Thereby, the range of the AMDF result for estimating the cycle (movement cycle) for one step of walking motion can be narrowed. Therefore, it is possible to estimate the cycle (motion cycle) for one step of walking motion with a smaller amount of calculation.

Further, the period determining unit 17 extracts the AMDF result in the extraction range described above, and further, among the local minimum values in the extraction range of the AMDF result, the difference from the minimum value of the extracted AMDF result is a preset constant c (first It is determined whether or not there is a minimum value that is less than or equal to (threshold of 2), and if there is no minimum value that is less than or equal to the constant c, the lag that takes the minimum value of the extracted AMDF result is taken as the walking cycle (motion cycle) ). In addition, when there is a minimum value that is equal to or less than the constant c, the cycle determination unit 17 determines the minimum value that is closest to the half value of the lag that takes the minimum value of the AMDF result among the minimum values that are equal to or less than the constant c. The lag is estimated as a walking cycle (motion cycle).
Thereby, even when there is much distortion in the acceleration data, the cycle (movement cycle) of one step of the walking motion can be estimated with high accuracy.

The present invention is not limited to the above embodiments, and can be modified without departing from the spirit of the present invention. In each of the above embodiments, the sensor data storage unit 12, the window division data storage unit 14, the AMDF storage unit 16, and the cycle storage unit 18 have been described as separate storage units, but the present invention is not limited thereto. is not. For example, the form which provides a some memory | storage part as one memory | storage part may be sufficient, and the form which provides a memory | storage part outside the portable terminal device 1 may be sufficient.
In each of the above embodiments, the sampling unit 11, the window dividing unit 13, the AMDF calculating unit 15, and the period determining unit 17 may be realized by dedicated hardware.

  In addition, the functions of the sampling unit 11, the window dividing unit 13, the AMDF calculating unit 15, the period determining unit 17 and the like described above are recorded on a computer-readable recording medium and recorded on the recording medium. Each process may be performed by causing the computer system to read and execute the program. Here, the “computer system” may include an OS and hardware such as peripheral devices. Further, the “computer system” includes a homepage providing environment (or display environment) if a WWW system is used. The “computer-readable recording medium” means a flexible disk, a magneto-optical disk, a ROM, a writable nonvolatile memory such as a flash memory, a portable medium such as a CD-ROM, a hard disk built in a computer system, etc. This is a storage device.

Further, the “computer-readable recording medium” means a volatile memory (for example, DRAM (Dynamic DRAM) in a computer system that becomes a server or a client when a program is transmitted through a network such as the Internet or a communication line such as a telephone line. Random Access Memory)), etc., which hold programs for a certain period of time. The program may be transmitted from a computer system storing the program in a storage device or the like to another computer system via a transmission medium or by a transmission wave in the transmission medium.
Here, the “transmission medium” for transmitting the program refers to a medium having a function of transmitting information, such as a network (communication network) such as the Internet or a communication line (communication line) such as a telephone line. The program may be for realizing a part of the functions described above. Furthermore, what can implement | achieve the function mentioned above in combination with the program already recorded on the computer system, and what is called a difference file (difference program) may be sufficient.

  As another application of the present invention, it is possible to estimate whether the subject is in a walking state or in a running state by making a determination using a threshold with respect to the cycle of the walking motion. The estimation accuracy and power consumption at that time are also superior to those of FFT. In addition, the technique is applicable not only to walking motions but also to a pair of human motions performed at a cycle of about several Hz to 10 Hz, for example, exercise such as swimming.

DESCRIPTION OF SYMBOLS 1 Portable terminal device 10 Acceleration measurement part 11 Sampling part 12 Sensor data memory | storage part 13 Window division | segmentation part 14 Window division | segmentation data memory | storage part 15 AMDF calculating part 16 AMDF memory | storage part 17 Period determination part 18 Period memory | storage part 19 Data transmission part 20 Power supply part

Claims (11)

A cycle estimation device that estimates a motion cycle indicating a period of one pair of motion motions with respect to acceleration data acquired by a mobile acceleration measurement unit,
An arithmetic unit for performing an AMDF (Average Magnitude Difference Function) operation on the acceleration data;
A period estimation unit for estimating the motion period based on a minimum value of the AMDF result calculated by the arithmetic unit and a minimum value in an extraction range of the AMDF result determined based on the minimum value. A period estimation device. The period estimation device includes:
A dividing unit that divides the acceleration data into divided data of a predetermined section;
The computing unit is
The period estimation apparatus according to claim 1, wherein an AMDF operation is performed on the divided data divided by the dividing unit. The period estimation unit includes:
Determining whether or not there is a minimum value in which the difference from the calculated minimum value of the AMDF result is equal to or less than a preset first threshold value among the minimum values in the calculated AMDF result;
When there is a minimum value that is less than or equal to the first threshold, among the lags corresponding to the minimum value that is less than or equal to the first threshold, a range less than or equal to the minimum lag is set as the extraction range,
When the minimum value which becomes below the said 1st threshold value does not exist, the range below the lag corresponding to the calculated minimum value of the AMDF result is set as the extraction range. Item 3. The period estimation device according to Item 2. The period estimation unit includes:
The AMDF result in the extraction range is extracted, and further, among the minimum values in the extraction range, whether or not there is a minimum value whose difference from the minimum value of the extracted AMDF result is equal to or smaller than a preset second threshold value. Determine whether
If there is no local minimum that is less than or equal to the second threshold, a lag that takes the minimum value of the extracted AMDF result is estimated as the motion period,
When there is a minimum value that is less than or equal to the second threshold, among the minimum values that are less than or equal to the second threshold, the lag of the minimum value closest to the half value of the lag that takes the minimum value of the AMDF result is determined. The period estimation apparatus according to any one of claims 1 to 3, wherein the period is estimated as the movement period. The motion cycle is
The cycle estimation device according to any one of claims 1 to 4, wherein the cycle is a walking cycle indicating a period of one step of walking motion. A cycle estimation method for estimating a motion cycle indicating a period of one pair of motion motions for acceleration data acquired by a portable acceleration measurement unit,
Performing AMDF (Average Magnitude Difference Function) operation on the acceleration data;
Estimating the motion period based on the minimum value of the AMDF result calculated by the arithmetic unit and the minimum value in the extraction range of the AMDF result determined based on the minimum value. Period estimation method. The period estimation method includes:
Dividing the acceleration data into divided data of a predetermined section,
The step of performing the AMDF operation includes:
The period estimation method according to claim 6, further comprising: performing an AMDF operation on the divided divided data. The step of estimating the movement period includes:
Determining whether there is a local minimum value in which the difference from the minimum value of the calculated AMDF result is equal to or less than a preset first threshold value among the local minimum values in the calculated AMDF result;
When there is a minimum value that is less than or equal to the first threshold, among the lags corresponding to the minimum value that is less than or equal to the first threshold, a range that is less than or equal to the minimum lag is set as the extraction range; ,
When there is no local minimum value that is less than or equal to the first threshold value, a range that is less than or equal to the lag corresponding to the calculated minimum value of the AMDF result is set as the extraction range. The period estimation method according to claim 6 or claim 7. The step of estimating the movement period includes:
Extracting an AMDF result in the extraction range;
Determining whether there is a local minimum value in which the difference from the minimum value of the extracted AMDF result is equal to or less than a preset second threshold value among the local minimum values in the extraction range;
Estimating a lag that takes the minimum value of the extracted AMDF result as the motion period if there is no local minimum that is less than or equal to the second threshold;
When there is a minimum value that is less than or equal to the second threshold, among the minimum values that are less than or equal to the second threshold, the lag of the minimum value closest to the half value of the lag that takes the minimum value of the AMDF result is determined. The period estimation method according to claim 6, further comprising a step of estimating the movement period. The motion cycle is
The cycle estimation method according to any one of claims 6 to 9, wherein the cycle is a walking cycle indicating a period of one walking step. To a computer as a cycle estimation device that estimates a motion cycle indicating a period of one pair of exercise motions for acceleration data acquired by a portable acceleration measurement unit,
A calculation unit that performs an AMDF (Average Magnitude Difference Function) operation on the acceleration data;
A period estimation procedure in which the period estimation unit estimates the motion period based on the minimum value of the AMDF result calculated by the calculation unit and the minimum value in the extraction range of the AMDF result determined based on the minimum value; A program for running

Images