JP2011188901A - Movement detector, movement detection method, and program - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a movement detector which accurately detects the movement. <P>SOLUTION: The movement detector detects the body movements, generates first frequency data by converting first time data showing detected time-series movements into data in a frequency region, generates second frequency data by converting the generated first frequency data non-linearly, generates second time data by converting the generated second frequency data into data in a time region, generates third time data by extracting low frequency components from the generated second time data, generates third frequency data by converting the generated third time data into data in a frequency region, and detects the pace of the movements from peak positions that appear in the generated third frequency data. <P>COPYRIGHT: (C)2011,JPO&INPIT

Description

  The present invention relates to a motion detection device, a motion detection method, and a program.

  Attempts have been made to analyze body movements and use the results of such analysis in health management and rehabilitation situations. A problem in proceeding with these attempts is how to accurately detect body movements. The movement of the body is detected by using a motion sensor such as an acceleration sensor, a gyro sensor, a skin temperature sensor, or a pulse sensor. For example, by using an acceleration sensor or a gyro sensor, it is possible to detect a daily action such as walking, running, sitting, standing, lying down.

  As a daily behavior detection method using a motion sensor, for example, Patent Literature 1 below discloses a method for detecting a daily behavior from a waveform of sensor data indicating a detection result by a motion sensor. Usually, in a sitting state or a lying state, the waveform of sensor data obtained by the acceleration sensor has a small amplitude. Therefore, when the amplitude of the sensor data is smaller than a predetermined threshold value, it can be determined that the user is sitting or lying down. Furthermore, it is possible to distinguish between a sitting state and a lying state by determining the direction in which gravity is applied. By using the method disclosed in this document, it is possible to detect daily behavior in this way.

  Patent Document 2 below discloses a method for detecting the pace of exercise rather than the type of daily action. This document discloses, for example, a method for detecting the pace of walking by analyzing the waveform of sensor data obtained by an acceleration sensor. Usually, walking is an exercise having a periodic pattern in which one step is one cycle. In particular, during walking, body parts such as the waist, legs, and arms repeat movement at a certain pace. Therefore, it is possible to detect the pace of walking by attaching acceleration sensors to these body parts and detecting the periodicity appearing in the waveform of the sensor data obtained by the acceleration sensor. By using the method disclosed in this document, the pace of walking can be detected in this way.

JP-A-10-24026 JP 2008-154733 A

  However, even if the methods disclosed in the above-mentioned documents are used, it is not possible to detect “movement” (for example, periodic movement such as walking) with sufficient accuracy. In general, sensor data includes a frequency component that directly reflects the motion of a certain body part and other frequency components. Other frequency components include, for example, fine vibrations generated by other body parts that are not directly related to exercise. The frequency component reflecting the movement appears in the low frequency band. On the other hand, other frequency components appear in the high frequency band. However, since all frequency components are generated due to the movement of the same living body, it is not easy to clearly separate them. The methods disclosed in the above documents cannot separate these frequency components with sufficient accuracy, and as a result, “motion” cannot be detected with sufficient accuracy.

  Therefore, the present invention has been made in view of the above problems, and an object of the present invention is to provide a new and improved motion detection apparatus, motion capable of detecting the pace of motion more accurately. To provide a detection method and a program.

  In order to solve the above-described problem, according to an aspect of the present invention, motion detection means for detecting body motion, and first time data indicating time-series motion detected by the motion detection means are represented in a frequency domain. First time / frequency conversion means for generating first frequency data by converting the first frequency data, and non-linear conversion of the first frequency data generated by the first time / frequency conversion means to obtain a second Non-linear conversion means for generating frequency data; first frequency / time conversion means for generating second time data by converting the second frequency data generated by the non-linear conversion means into time-domain data; A low frequency component extracting means for extracting a low frequency component from the second time data generated by the first frequency / time converting means and generating a third time data; and the low frequency component extracting means. Second time / frequency conversion means for converting the generated third time data into frequency domain data to generate third frequency data; and third time generated by the second time / frequency conversion means. And a pace detecting means for detecting the pace of the movement from a peak position appearing in the frequency data.

  The first time / frequency conversion means converts the first time data cut out at a predetermined sampling period and a predetermined cut-out interval from the detection result of the motion detection means into frequency domain data. You may be comprised so that 1st frequency data may be produced | generated. In this case, the low frequency component extracting means extracts a low frequency component from the second time data using a predetermined threshold corresponding to the predetermined sampling period and the predetermined cut-out interval.

  The pace detecting means interpolates the third frequency data, which is a discrete data string, by a predetermined interpolation method, and detects the pace of movement from a peak position appearing in the third frequency data after interpolation. It may be configured to.

  Further, the motion detection device generates fourth frequency data by performing inverse transformation of nonlinear transformation by the nonlinear transformation unit on the third frequency data generated by the second time / frequency transformation unit. Non-linear inverse transforming means, second frequency / time transforming means for transforming the fourth frequency data generated by the nonlinear inverse transforming means into time domain data to generate fourth time data, Data storage means for storing the fourth time data generated in advance for each type, and a fourth time generated by the second frequency / time conversion means based on the motion detected by the motion detection means The exercise type determination means for comparing the data and the fourth time data stored in the data storage means to determine the type of exercise corresponding to the movement detected by the movement detection means , It may further include a.

The exercise type determination means includes fourth time data generated by the second frequency / time conversion means based on the movement detected by the movement detection means, and a fourth time data stored in the data storage means. Similarity calculation means for calculating the pattern similarity between the time data of 4 and
Of the fourth time data stored in the data storage means, fourth time data having a high pattern similarity calculated by the similarity calculation means is extracted, and an exercise corresponding to the extracted fourth time data And a determination result output means for outputting the type.

  In order to solve the above-described problem, according to another aspect of the present invention, a motion detection step for detecting body motion, and first time data indicating time-series motion detected in the motion detection step A first time / frequency conversion step for generating first frequency data by converting the data into frequency domain data, and a non-linear conversion of the first frequency data generated in the first time / frequency conversion step. Non-linear conversion step for generating second frequency data, and first frequency / time conversion for generating second time data by converting the second frequency data generated in the non-linear conversion step into time domain data A low frequency component extracting step for extracting a low frequency component from the second time data generated in the first frequency / time conversion step and generating a third time data; A second time / frequency conversion step of generating third frequency data by converting the third time data generated in the low frequency component extraction step into frequency domain data; and the second time / frequency conversion step. And a pace detecting step of detecting the pace of the movement from a peak position appearing in the third frequency data generated in step (3).

  In order to solve the above-described problem, according to another aspect of the present invention, a motion detection function for detecting body motion, and first time data indicating time-series motion detected by the motion detection function A first time / frequency conversion function for generating first frequency data by converting the data into frequency domain data, and a non-linear conversion of the first frequency data generated by the first time / frequency conversion function Non-linear conversion function for generating second frequency data, and first frequency / time conversion for generating second time data by converting the second frequency data generated by the non-linear conversion function into time-domain data A low frequency component extracting function for generating a third time data by extracting a low frequency component from the second time data generated by the first frequency / time converting function, and the low frequency component extracting function The third time data generated by the second time / frequency conversion function for converting the third time data generated into frequency domain data to generate the third frequency data, and the second time / frequency conversion function generated by the second time / frequency conversion function. And a pace detecting function for detecting the pace of the movement from the peak position appearing in the frequency data of 3 is provided.

  Furthermore, in order to solve the above-described problem, according to another aspect of the present invention, a computer-readable recording medium on which the program is recorded is provided.

  As described above, according to the present invention, the pace of exercise can be detected with higher accuracy.

It is explanatory drawing for demonstrating the function structure of the sensing apparatus which concerns on 1st Embodiment of this invention, and a motion analysis apparatus. It is explanatory drawing for demonstrating the detailed function structure of the data processing means contained in the exercise | movement analyzer which concerns on the embodiment. It is explanatory drawing for demonstrating the waveform of sensor data. It is explanatory drawing for demonstrating the shape of a spectrum. It is explanatory drawing for demonstrating the shape of a logarithmic spectrum. It is explanatory drawing for demonstrating the spectrum shape after high-pass filtering. It is explanatory drawing for demonstrating the exercise | movement detection method which concerns on the same embodiment. It is explanatory drawing for demonstrating the detailed function structure of the data processing means contained in the exercise | movement analyzer which concerns on 2nd Embodiment of this invention. It is explanatory drawing for demonstrating the exercise | movement detection method which concerns on the same embodiment. It is explanatory drawing for demonstrating the hardware constitutions which can implement | achieve the function of the motion analysis apparatus which concerns on 1st and 2nd embodiment of this invention.

  Exemplary embodiments of the present invention will be described below in detail with reference to the accompanying drawings. In addition, in this specification and drawing, about the component which has the substantially same function structure, duplication description is abbreviate | omitted by attaching | subjecting the same code | symbol.

[About the flow of explanation]
Here, the flow of explanation regarding the embodiment of the present invention described below will be briefly described. First, a system configuration for realizing the motion detection method according to the first embodiment of the present invention will be described with reference to FIG. In the description, the configurations of the sensing device 10 and the motion analysis device 20 according to the embodiment will be described. Next, a detailed configuration of the data processing unit 203 included in the motion analysis apparatus 20 of the embodiment will be described with reference to FIG. Next, the motion detection method according to the embodiment will be described with reference to FIG.

  Next, a detailed configuration of the data processing unit 203 included in the motion analysis apparatus 20 according to the second embodiment of the present invention will be described with reference to FIG. Next, a motion detection method according to the embodiment will be described with reference to FIG. Next, a hardware configuration capable of realizing the functions of the motion analysis apparatus 20 according to the first and second embodiments of the present invention will be described with reference to FIG.

<1: First Embodiment>
A first embodiment of the present invention will be described. The present embodiment relates to a method for accurately detecting the pace of movement from sensor data obtained by a movement sensor.

[1-1: Introduction]
First, features of sensor data obtained by the motion sensor will be described. As an example, measured values of sensor data obtained by an acceleration sensor attached to the upper arm by operating the weight machine four times are shown in FIG. Referring to the sensor data in FIG. 3, it can be seen that there are a large amplitude portion and a small amplitude portion in the waveform. Moreover, it turns out that the part with a large amplitude and the small part are repeated alternately. Thus, the sensor data includes a characteristic pattern that reflects the motion of the exercise.

  Such a pattern appears as a rough shape corresponding to a waveform envelope (hereinafter referred to as a spectral envelope component). However, as can be seen from the example of FIG. 3, the waveform of the sensor data includes a fine vibration component (hereinafter, spectral fine structure). Therefore, in order to analyze the characteristic pattern appearing in the waveform of the sensor data with high accuracy, the spectral fine structure and the spectral envelope component must be separated. However, in the conventional method, the motion analysis is not performed after the spectral fine structure and the spectral envelope component are sufficiently separated, and the pace of the motion cannot be detected with sufficient accuracy. Therefore, the present inventor has devised a method of sufficiently removing the spectral fine structure that hinders the detection of the pace of movement and detecting the pace of movement more accurately. Hereinafter, this method will be described in detail.

[1-2: System configuration]
First, a system configuration according to the present embodiment will be described with reference to FIG. In particular, the configuration of the sensing device 10 and the configuration of the motion analysis device 20 according to the present embodiment will be described. FIG. 1 is an explanatory diagram for explaining a configuration of a sensing device 10 and a motion analysis device 20 according to the present embodiment.

  This system includes a sensing device 10 and a motion analysis device 20. The sensing device 10 provides a function of a motion sensor. For example, the sensing device 10 is used by being attached to a body part such as an upper arm, a wrist, or a waist. On the other hand, the exercise analysis device 20 is used to detect the pace of exercise by analyzing sensor data acquired by the sensing device 10. The sensing device 10 and the motion analysis device 20 are connected by a wired or wireless communication path. The sensor data acquired by the sensing device 10 is transmitted to the motion analysis device 20 through a wired or wireless communication path. As a wireless communication path, for example, Bluetooth (registered trademark), ZigBee (registered trademark), wireless LAN, or the like can be applied.

(Configuration of sensing device 10)
As shown in FIG. 1, the sensing device 10 mainly includes a sensing unit 101 and a data transmission unit 102.

  The sensing unit 101 is, for example, a motion sensor such as an acceleration sensor, a gyro sensor, a skin temperature sensor, a pulse sensor, an angular velocity sensor, a myoelectric potential sensor, or a tilt sensor. In the following, an acceleration sensor is assumed as the sensing means 101 for convenience of explanation. Of course, the sensing means 101 is not limited to an acceleration sensor. When the user wearing the sensing device 10 exercises, the sensing means 101 detects the exercise and outputs sensor data. For example, the sensing means 101 detects acceleration generated due to movement and outputs the acceleration data as sensor data. The sensor data output from the sensing unit 101 is input to the data transmission unit 102. When sensor data is input, the data transmission unit 102 transmits the input sensor data to the motion analysis device 20 through a wired or wireless communication path.

(Configuration of the motion analysis device 20)
On the other hand, as shown in FIG. 1, the motion analysis apparatus 20 mainly includes a data receiving unit 201, a data storage unit 202, a data processing unit 203, and a display unit 204.

  The sensor data transmitted from the sensing device 10 is received by the data receiving unit 201. The sensor data received by the data receiving unit 201 is stored in the data storage unit 202. The sensor data stored in the data storage unit 202 is read out by the data processing unit 203. Then, the data processing unit 203 analyzes the sensor data read from the data storage unit 202 and detects the pace of exercise. The configuration of the data processing unit 203 will be described in detail later. When the pace of exercise is detected by the data processing unit 203, the detection result is input to the display unit 204. The display unit 204 displays the pace of exercise detected by the data processing unit 203.

  As described above, sensor data is acquired by the sensing device 10, and the pace of exercise is detected from the sensor data by the motion analysis device 20. The present embodiment is characterized by a method for analyzing sensor data by the data processing unit 203. Therefore, the configuration of the data processing unit 203 will be described in detail below.

[1-3: Details of Data Processing Unit 203]
As shown in FIG. 2, the data processing unit 203 mainly includes a first frequency conversion unit 2031, a logarithmic conversion unit 2032, a frequency inverse conversion unit 2033, a low-pass filter 2034, and a second frequency conversion unit 2035. And pace detecting means 2036.

  As described above, the data processing unit 203 reads the sensor data stored in the data storage unit 202 and analyzes the read sensor data to detect the pace of exercise. The data storage unit 202 stores sensor data transmitted from the sensing device 10. This sensor data is time-series data (see, for example, FIG. 3) detected by the sensing means 101 at a predetermined sampling period (for example, a sampling frequency of 50 Hz). When reading the sensor data stored in the data storage unit 202, the data processing unit 203 cuts out sensor data to be analyzed at a predetermined time interval (cutout section).

  Note that when the type of exercise is switched and the start point and end point of the exercise can be read from the sensor data, the data processing unit 203 sets the interval from the start point to the end point of the exercise as the analysis target segment. To extract sensor data. On the other hand, when the start point and end point of the exercise cannot be read from the sensor data, the data processing unit 203 cuts out the sensor data to be analyzed based on an arbitrarily set cut-out section (for example, 2 seconds, 4 seconds, etc.). When the sensing means 101 of the sensing device 10 is a 2-axis acceleration sensor, a 3-axis acceleration sensor, or the like, it is necessary to synthesize sensor data detected for each axis. It is executed at the read timing.

  The sensor data read from the data storage unit 202 in this way is input to the first frequency conversion unit 2031. Since the sensor data input to the first frequency conversion means 2031 is time domain data, this data is hereinafter referred to as first time data. When the first time data is input, the first frequency conversion means 2031 performs a predetermined time-frequency conversion on the input first time data, and data in the frequency domain (hereinafter referred to as first frequency data). Is calculated. As the predetermined time / frequency transform, for example, Fourier transform, Hadamard transform, cosine transform, Hilbert transform, discrete wavelet transform, or the like can be applied. In the following description, Fourier transform is assumed as the predetermined time / frequency conversion.

For example, the first frequency data obtained by Fourier transforming the first time data has a waveform as shown in FIG. FIG. 4 is first frequency data (spectral intensity) obtained by Fourier transforming the first time data shown in FIG. By this Fourier transform, a spectral intensity up to the Nyquist frequency (1/2 of the sampling frequency) can be obtained. If the first time data is x _k (k = 0,..., N−1) and the first frequency data is X _n (n = 0,..., N−1), the first frequency data X _n is obtained by the following formula (1). Note that | X _n | ² is referred to as spectral intensity. Furthermore, when analyzing the first frequency data X _n, in order to eliminate the influence of random noise occurring, such as during the measurement of the sensor data, using a periodogram obtained by averaging the plurality of first frequency data X _n Is preferred.

…（１）

... (1)

The first frequency data _Xn calculated by the first frequency conversion unit 2031 in this way is input to the logarithmic conversion unit 2032. When the first frequency data is input, the logarithmic conversion unit 2032 performs logarithmic conversion on the spectrum intensity | X _n | ² of the input first frequency data to obtain the second frequency data (logarithmic spectrum intensity: log). | X _n | ² ) is calculated. As described above, the sensor data includes a low frequency component directly related to the pace of movement and a high frequency component resulting from other movements. Now, when the first component of the frequency data _{X n L n} corresponding to the low-frequency component, a component corresponding to the high frequency component and _{H n,} log spectral magnitudes log _| X n ^{| 2} is the following formula (2) It is expanded like this.

…（２）

... (2)

As can be seen from the above equation (2), by performing logarithmic transformation, the low frequency component and the high frequency component of the sensor data included in the first frequency data in the form of convolution operation are converted into the form of sum operation. Converted. Thus, by converting from the form of convolution operation to the form of sum operation, it becomes easy to separate the low frequency component and the high frequency component (see FIG. 5). Here, logarithmic transformation is taken as an example, but other nonlinear transformation methods that can obtain the same effect as logarithmic transformation can also be applied. As another nonlinear conversion method, for example, “negative number of inverse conversion of spectrum intensity (−1 / | X _n | ² )”, “square root conversion”, or the like can be applied.

  The second frequency data calculated by the logarithmic conversion unit 2032 in this way is input to the frequency inverse conversion unit 2033. When the second frequency data is input, the frequency inverse conversion means 2033 performs a predetermined frequency / time conversion on the input second frequency data to calculate the second time data. The predetermined frequency / time conversion used here is an inverse conversion of the predetermined time / frequency conversion by the first frequency conversion means 2031. Therefore, the type of the predetermined frequency / time conversion used here is also determined according to the type of the predetermined time / frequency conversion used. However, here, it is assumed that the inverse Fourier transform (see Equation (3) below) is performed on the second frequency data, and the second time data is calculated.

…（３）

... (3)

  The second time data calculated by the frequency inverse transform unit 2033 in this way is input to the low-pass filter 2034. When the second time data is input, the low-pass filter 2034 cuts the high frequency component included in the input second time data and extracts the low frequency component. Usually, the band of the low frequency component and the band of the high frequency component generated by motion are close. Therefore, when the logarithmic conversion is not performed by the logarithmic conversion means 2032, it is difficult to accurately cut only the high frequency component by the low-pass filter 2034. However, the second time data input to the low-pass filter 2034 has been logarithmically converted, and the low-frequency component and the high-frequency component are separated on the frequency axis. Only high frequency components can be easily cut.

  Here, the relationship between the intensity of exercise and the high frequency component will be described. Generally, when exercising with a large acceleration, a load (resistance) applied to a body part that moves is reduced. Therefore, fine vibrations (corresponding to high-frequency components) generated in the body during exercise become relatively coarse vibrations (vibrations having a relatively low frequency). On the other hand, when exercising with a small acceleration (corresponding to a low-frequency component), the load (resistance) applied to the moving body part increases. Therefore, fine vibrations (corresponding to high frequency components) generated in the body during exercise become relatively fine vibrations (vibrations having a relatively high frequency).

When the intensity of motion is large, the maximum value | X _max | ² of the spectral intensity becomes large. In this case, the peak of the high frequency component appears in a relatively low range. On the other hand, when the intensity of the motion is small, the maximum value | X _max | ² of the spectrum intensity becomes small. In this case, the peak of the high frequency component appears in a relatively high range. That is, the position where the peak of the high frequency component appears changes according to the magnitude of the maximum value | X _max | ² of the spectrum intensity. Therefore, the threshold value q used to cut the high-frequency component in the low-pass filter 2034 is set to a value corresponding to the maximum value | X _max | ² of the spectrum intensity (see the following formula (4)).

…（４）

... (4)

However, in the above equation (4), the parameter α is a constant determined according to the sampling frequency and the length of the cut-out section. As the parameter α, a value measured in advance by a prior experiment is used. For example, n sensor data are measured as a preliminary experiment, and the spectral intensity | X _max | ² is detected for each of them, and then the valley of the spectrum shown in FIG. 5 is determined. This valley of the spectrum is located in a lower region than the peak of the high frequency component, and indicates a position on the frequency axis where the logarithmic spectrum intensity is minimum.

The valley of this spectrum is detected for n sensor data and substituted for q in the above equation (4). Further, the spectral intensity | X _max | ² detected for n sensor data is substituted into the above equation (4). Then, using the n parameters α obtained by the above equation (4), the value of the parameter α that minimizes the sum of the square errors is estimated using regression analysis or the like. By applying such a method, the parameter α included in the above equation (4) can be obtained. For example, when the sampling frequency is 50 Hz and the cut-out section is 4 seconds, it is estimated that α = 2.1 × 10 ⁻¹⁷ .

  Data from which the high-frequency component is cut by the low-pass filter 2034 using the threshold value q (hereinafter, third time data) is input to the second frequency conversion means 2035. When the third time data is input, the second frequency conversion means 2035 performs time / frequency conversion on the input third time data to calculate the third frequency data. The predetermined time / frequency conversion used here is the same as the predetermined time / frequency conversion by the first frequency conversion means 2031. Here, Fourier transform is performed on the third time data, and the third frequency data is calculated.

  The third frequency data calculated by the second frequency conversion unit 2035 in this way is input to the pace detection unit 2036. When the third frequency data is input, the pace detecting means 2036 detects a peak appearing in the input third frequency data, and outputs the frequency at which the detected peak is located as the pace of exercise. At this time, the pace detection means 2036 performs predetermined linear or nonlinear interpolation on the third frequency data which is a discrete data string, and detects a peak from the third frequency data after the interpolation (see FIG. 6). . The exercise pace output from the pace detection means 2035 is input to the display means 204.

  Here, a supplementary explanation will be given regarding the necessity of interpolation. Generally, the pace of exercise using a weight machine is about 0.1 to 0.5 Hz. For example, if the cutout section is 4 seconds, the frequency resolution is 0.25 Hz (reciprocal of 4 seconds). That is, with a resolution of about 0.25 Hz, an exercise pace of about 0.1 to 0.5 Hz cannot be detected with sufficient accuracy. Such a problem does not occur when analyzing a signal of about 1 kHz. However, when analyzing an exercise pace with a low frequency, it is necessary to devise a method for compensating for the low resolution so that the exercise pace can be detected with sufficient accuracy.

  As a contrivance, the present inventors have devised a method of interpolating the third frequency data obtained by a discrete data string. As an interpolation method, for example, spline interpolation, Lagrangian interpolation, Hermite interpolation, or the like is applicable. In the case of cubic spline interpolation, the third frequency data can be interpolated by the following method.

When a set of frequency and spectrum intensity is (f ₀ , p ₀ ),..., (F _N , p _N ), and an interpolation function in the interval [f _j , f _{j + 1} ] is S _j (x), this interpolation function S _j (x) are unknown constants _{a j, b j, c j} , with _{d j,} is expressed as the following equation (5). Since there are four unknown constants, 4 × N equations are established to determine these unknown constants. The following equations (6) to (9) are obtained from the boundary conditions. By solving these 4 × N equations, four unknown constants are calculated, and a cubic spline function S _j (x) is obtained. By performing such interpolation, the pace of movement is detected with high accuracy by compensating for the lack of resolution.

（ｊ＝０，１，２，…，Ｎ−１）
…（５）

…（６）

（Ｓ’はＳの一次導関数）
…（７）

（Ｓ”はＳの二次導関数）
…（８）

（Ｓ”はＳの二次導関数）
…（９）

(J = 0, 1, 2, ..., N-1)
... (5)

(6)

(S 'is the first derivative of S)
... (7)

(S ″ is the second derivative of S)
(8)

(S ″ is the second derivative of S)
... (9)

  The detailed configuration of the data processing unit 203 has been described above.

[1-4: Motion detection method]
Next, the motion detection method according to the present embodiment will be described with reference to FIG. FIG. 7 is an explanatory diagram showing a flow of processing for realizing the motion detection method according to the present embodiment. Here, the operation flow of the sensing device 10 and the motion analysis device 20 will be described.

  First, motion is detected by the sensing device 10 and sensor data is acquired (S101). Next, sensor data is transmitted from the sensing device 10 to the motion analysis device 20, and the sensor data is cut out by the motion analysis device 20 (S102). At this stage, the sensor data is preprocessed into a format that can be processed by the data processing means 203. Next, the data processing unit 203 performs time / frequency conversion on the sensor data (first time data) to generate first frequency data (S103).

  Next, the data processing unit 203 performs logarithmic transformation on the spectrum intensity of the first frequency data to generate second frequency data (S104). Next, the data processing unit 203 performs frequency / time conversion on the second frequency data to generate second time data (S105). Next, the data processing unit 203 performs high-pass filtering on the second time data to cut high frequency components, and generates third time data (S106).

  Next, the data processing unit 203 performs time / frequency conversion on the third time data to generate third frequency data (S107). Next, the data processing unit 203 performs spline interpolation on the third frequency data, and detects the pace of movement from the third frequency data after the interpolation (S108). Next, the display unit 204 displays the pace of exercise detected by the data processing unit 203 (S109). The process is executed according to the flow described above, and the motion detection method according to the present embodiment is realized.

  The first embodiment of the present invention has been described above.

<2: Second Embodiment>
A second embodiment of the present invention will be described. The present embodiment relates to a motion detection method in which the sensor data analysis method according to the first embodiment is applied to determination of a motion type. The main difference from the first embodiment is the configuration of the data processing unit 203. For this reason, the following description will be omitted, and only the differences will be described in detail.

[2-1: Details of Data Processing Unit 203]
As shown in FIG. 8, the data processing means 203 according to this embodiment mainly includes a first frequency conversion means 2131, a logarithmic conversion means 2132, a first frequency inverse conversion means 2133, a low-pass filter 2134, The second frequency converting means 2135, the exponent converting means 2136, the second frequency inverse converting means 2137, the motion determining means 2138, and the teacher data accumulating means 2139.

  As in the first embodiment, the sensor data read from the data storage unit 202 is input to the first frequency conversion unit 2131. The sensor data input to the first frequency conversion unit 2131 is referred to as first time data. When the first time data is input, the first frequency conversion means 2131 performs a predetermined time-frequency conversion on the input first time data to obtain frequency domain data (hereinafter referred to as first frequency data). Is calculated. As the predetermined time / frequency transform, for example, Fourier transform, Hadamard transform, cosine transform, Hilbert transform, discrete wavelet transform, or the like can be applied. In the following description, Fourier transform is assumed as the predetermined time / frequency conversion.

The first frequency data obtained by Fourier transforming the first time data is input to the logarithmic conversion means 2132. When the first frequency data is input, the logarithmic conversion unit 2132 performs logarithmic conversion on the spectrum intensity | X _n | ² of the input first frequency data to obtain the second frequency data (logarithmic spectrum intensity: log). | X _n | ² ) is calculated. Here, logarithmic transformation is taken as an example, but other nonlinear transformation methods that can obtain the same effect as logarithmic transformation can also be applied. As another nonlinear conversion method, for example, “negative number of inverse conversion of spectrum intensity (−1 / | X _n | ² )”, “square root conversion”, or the like can be applied.

  The second frequency data calculated by the logarithmic conversion unit 2132 in this way is input to the first frequency inverse conversion unit 2133. When the second frequency data is input, the first frequency reverse conversion unit 2133 performs predetermined frequency / time conversion on the input second frequency data to calculate second time data. The predetermined frequency / time conversion used here is an inverse conversion of the predetermined time / frequency conversion by the first frequency conversion means 2131.

  The second time data calculated by the first frequency inverse transform unit 2133 in this way is input to the low pass filter 2134. When the second time data is input, the low-pass filter 2134 cuts the high frequency component included in the input second time data and extracts the low frequency component. The configuration of the low-pass filter 2134 is substantially the same as the configuration of the low-pass filter 2034 according to the first embodiment. That is, using the threshold value q described above, data (hereinafter referred to as third time data) from which high-frequency components have been cut by the low-pass filter 2134 is obtained, and the data is input to the second frequency conversion means 2135. The

  When the third time data is input, the second frequency conversion means 2135 performs time / frequency conversion on the input third time data to calculate the third frequency data. The predetermined time / frequency conversion used here is the same as the predetermined time / frequency conversion by the first frequency conversion means 2131. The third frequency data calculated by the second frequency converting means 2135 in this way is input to the exponent converting means 2136. When the third frequency data is input, the exponent conversion unit 2136 performs exponential conversion on the input third frequency data to generate fourth frequency data.

  This exponential conversion is an inverse conversion of the logarithmic conversion by the logarithmic conversion means 2132. Therefore, when other nonlinear transformation is used instead of logarithmic transformation, the inverse transformation of other nonlinear transformation is applied instead of exponential transformation. The fourth frequency data generated by the exponent converting unit 2136 in this way is input to the second frequency inverse converting unit 2137. When the fourth frequency data is input, the second frequency inverse transform unit 2137 performs predetermined frequency / time conversion on the input fourth frequency data to generate fourth time data. The predetermined frequency / time conversion used here is an inverse conversion of the predetermined time / frequency conversion by the first frequency conversion means 2131.

  In this way, the fourth time data generated by the second frequency inverse transform unit 2137 is input to the motion determination unit 2138. When the fourth time data is input, the exercise determination unit 2138 compares the input fourth time data with the teacher data stored in advance in the teacher data storage unit 2139 to determine the type of exercise.

  Here, the teacher data stored in advance in the teacher data storage unit 2139 will be described. The teacher data referred to here is characteristic data whose type of exercise is known in advance. For example, sensor data detected by the sensing device 10 during walking, or data generated by processing this sensor data is characteristic data reflecting the type of motion “walking”. In this way, characteristic data detected and processed in advance for each type of exercise is called teacher data. Teacher data storage means 2139 stores teacher data generated in advance for various types of exercise.

  Here, high-frequency components are cut by the low-pass filter 2134, and characteristic data that strongly reflects the spectrum envelope component of the sensor data that characterizes the type of exercise is used as teacher data. Specifically, for example, sensor data detected by the sensing device 10 for each predetermined exercise type is converted into first frequency conversion means 2131, logarithmic conversion means 2132, first frequency inverse conversion means 2133, low-pass filter 2134, first Data (fourth time data) processed by the two-frequency conversion unit 2135, the exponent conversion unit 2136, and the second frequency inverse conversion unit 2137 is stored in the teacher data storage unit 2139.

  The fourth time data corresponds to sensor data in which the spectral fine structure is suppressed. Therefore, by using the fourth time data whose type of exercise is known in advance as the teacher data and comparing this teacher data with the fourth time data obtained from the actually measured sensor data, higher accuracy can be obtained. It becomes possible to determine the exercise type. Note that not only the fourth time data itself, but also the pace of movement obtained by the movement detection method of the first embodiment described above, the window average value of the fourth time data (each component cut out in a predetermined length section) Average value), variance, peak intensity, peak position, etc. may be used as teacher data.

  When the fourth time data is obtained for the actually measured sensor data, and the fourth time data is input, the motion determination means 2138 has the input fourth time data (hereinafter, actually measured data) and The teacher data read from the teacher data storage unit 2139 is compared. As described above, the elements of the data to be compared include, for example, the value (corresponding to the waveform of the fourth time data) at each time point of the fourth time data, the pace of exercise, and the like. The motion determination means 2138 arranges these elements and expresses them as feature quantity vectors, and calculates the similarity (hereinafter referred to as pattern similarity) between the feature quantity vector of the actually measured data and the feature quantity vector of the teacher data.

  As described above, the teacher data storage unit 2139 stores teacher data for each type of exercise. Therefore, the exercise determination unit 2138 calculates the pattern similarity for each exercise type, detects teacher data having a high pattern similarity, and determines the exercise type corresponding to the actual measurement data from the detection result. The comparison of the pattern similarity and the determination of the exercise type are realized by using a pattern recognizer based on a support vector machine or a hidden Markov model, for example.

  Here, let us consider a method for comparing the waveform patterns of the fourth time data. In this case, first, teacher data corresponding to the same time interval as the actually measured data cut out at a predetermined time interval is prepared. Also, the actual measurement data and the teacher data are normalized in the time interval. For example, the data value at each time is divided by the total data value in that time interval. By performing such preprocessing, it becomes possible to compare the waveform of the actual measurement data with the waveform of the teacher data regardless of the intensity of the sensor data that varies for each measurement.

Assuming that the actual measurement data at time k normalized in this way is f _k and the teacher data at time k corresponding to the exercise type Φ is f _kΦ , the difference between the actual measurement data and the teacher data at time k is {f _k − f _kΦ }. The exercise determination unit 2138 calculates a parameter _QΦ given by the following equation (10) for the teacher data corresponding to all exercise types Φ stored in the teacher data storage unit 2139. Then, the exercise determination unit 2138 detects the exercise type Φ that minimizes the parameter Q _Φ, and determines the detected exercise type Φ as the exercise type of the actually measured data.

…（１０）

(10)

  The exercise type of the actual measurement data determined by the exercise determination unit 2138 in this way is input to the display unit 204. When the exercise type of the actual measurement data is input, the display unit 204 displays the exercise type of the input actual measurement data. At this time, when the pace of exercise is detected by the data processing unit 203 as in the first embodiment, the display unit 204 may be configured to display the pace of exercise together. .

  The configuration of the data processing unit 203 according to the second embodiment of the present invention has been described above.

[2-2: Motion detection method]
Next, the motion detection method according to the present embodiment will be described with reference to FIG. FIG. 9 is an explanatory diagram showing a flow of processing for realizing the motion detection method according to the present embodiment. Here, the operation flow of the sensing device 10 and the motion analysis device 20 will be described.

  First, motion is detected by the sensing device 10 and sensor data is acquired (S201). Next, sensor data is transmitted from the sensing device 10 to the motion analysis device 20, and the sensor data is cut out by the motion analysis device 20 (S202). At this stage, the sensor data is preprocessed into a format that can be processed by the data processing means 203. Next, the data processing unit 203 performs time / frequency conversion on the sensor data (first time data) to generate first frequency data (S203).

  Next, the data processing unit 203 performs logarithmic transformation on the spectrum intensity of the first frequency data to generate second frequency data (S204). Next, the data processing unit 203 performs frequency / time conversion on the second frequency data to generate second time data (S205). Next, the data processing unit 203 performs high-pass filtering on the second time data to cut high frequency components, and generates third time data (S206).

  Next, the data processing unit 203 performs time / frequency conversion on the third time data to generate third frequency data (S207). Next, the data processing unit 203 performs exponential conversion on the third frequency data to generate fourth frequency data (S208). Next, the data processing unit 203 performs frequency / time conversion on the fourth frequency data to generate fourth time data (S209).

  Next, the data processing unit 203 compares the fourth time data obtained from the actually measured sensor data with the fourth time data corresponding to the sensor data measured in advance for each exercise type, to determine the type of exercise. Is determined (S210). Next, the display unit 204 displays the determination result of the exercise type by the data processing unit 203 (S211). At this time, it may be configured such that the pace of exercise detected in the same manner as the exercise detection method according to the first embodiment is displayed on the display unit 204.

  The process is executed according to the flow described above, and the motion detection method according to the present embodiment is realized.

<3: Hardware configuration>
The function of each component included in the motion analysis apparatus 20 can be realized using, for example, the hardware configuration of the information processing apparatus illustrated in FIG. That is, the function of each component is realized by controlling the hardware shown in FIG. 10 using a computer program. The form of the hardware is arbitrary, and includes, for example, a personal computer, a mobile phone, a portable information terminal such as a PHS, a PDA, a game machine, or various information appliances. However, the above PHS is an abbreviation of Personal Handy-phone System. The PDA is an abbreviation for Personal Digital Assistant.

  As shown in FIG. 10, this hardware mainly includes a CPU 902, a ROM 904, a RAM 906, a host bus 908, and a bridge 910. Further, this hardware includes an external bus 912, an interface 914, an input unit 916, an output unit 918, a storage unit 920, a drive 922, a connection port 924, and a communication unit 926. However, the CPU is an abbreviation for Central Processing Unit. The ROM is an abbreviation for Read Only Memory. The RAM is an abbreviation for Random Access Memory.

  The CPU 902 functions as, for example, an arithmetic processing unit or a control unit, and controls the overall operation of each component or a part thereof based on various programs recorded in the ROM 904, the RAM 906, the storage unit 920, or the removable recording medium 928. . The ROM 904 is a means for storing a program read by the CPU 902, data used for calculation, and the like. In the RAM 906, for example, a program read by the CPU 902, various parameters that change as appropriate when the program is executed, and the like are temporarily or permanently stored.

  These components are connected to each other via, for example, a host bus 908 capable of high-speed data transmission. On the other hand, the host bus 908 is connected to an external bus 912 having a relatively low data transmission speed via a bridge 910, for example. As the input unit 916, for example, a mouse, a keyboard, a touch panel, a button, a switch, a lever, or the like is used. Furthermore, as the input unit 916, a remote controller capable of transmitting a control signal using infrared rays or other radio waves may be used.

  As the output unit 918, for example, a display device such as a CRT, LCD, PDP, or ELD, an audio output device such as a speaker or a headphone, a printer, a mobile phone, or a facsimile, etc. Or it is an apparatus which can notify audibly. However, the above CRT is an abbreviation for Cathode Ray Tube. The LCD is an abbreviation for Liquid Crystal Display. The PDP is an abbreviation for Plasma Display Panel. Furthermore, the ELD is an abbreviation for Electro-Luminescence Display.

  The storage unit 920 is a device for storing various data. As the storage unit 920, for example, a magnetic storage device such as a hard disk drive (HDD), a semiconductor storage device, an optical storage device, a magneto-optical storage device, or the like is used. However, the HDD is an abbreviation for Hard Disk Drive.

  The drive 922 is a device that reads information recorded on a removable recording medium 928 such as a magnetic disk, an optical disk, a magneto-optical disk, or a semiconductor memory, or writes information to the removable recording medium 928. The removable recording medium 928 is, for example, a DVD medium, a Blu-ray medium, an HD DVD medium, or various semiconductor storage media. Of course, the removable recording medium 928 may be, for example, an IC card on which a non-contact type IC chip is mounted, an electronic device, or the like. However, the above IC is an abbreviation for Integrated Circuit.

  The connection port 924 is a port for connecting an external connection device 930 such as a USB port, an IEEE 1394 port, a SCSI, an RS-232C port, or an optical audio terminal. The external connection device 930 is, for example, a printer, a portable music player, a digital camera, a digital video camera, or an IC recorder. However, the above USB is an abbreviation for Universal Serial Bus. The SCSI is an abbreviation for Small Computer System Interface.

  The communication unit 926 is a communication device for connecting to the network 932. For example, a wired or wireless LAN, Bluetooth (registered trademark), or a WUSB communication card, an optical communication router, an ADSL router, or various types It is a modem for communication. The network 932 connected to the communication unit 926 is configured by a wired or wireless network, such as the Internet, home LAN, infrared communication, visible light communication, broadcast, or satellite communication. However, the above LAN is an abbreviation for Local Area Network. The WUSB is an abbreviation for Wireless USB. The above ADSL is an abbreviation for Asymmetric Digital Subscriber Line.

  As mentioned above, although preferred embodiment of this invention was described referring an accompanying drawing, it cannot be overemphasized that this invention is not limited to the example which concerns. It will be apparent to those skilled in the art that various changes and modifications can be made within the scope of the claims, and these are naturally within the technical scope of the present invention. Understood.

  For example, a removable storage medium is mounted on the sensing device 10, sensor data is stored in the removable storage medium, the removable storage medium is mounted on the motion analysis device 20, and the sensor data of the removable storage medium is analyzed. It may be configured as follows. With such a configuration, it is not necessary to provide a communication path between the sensing device 10 and the motion analysis device 20, and the degree of freedom of installation of the sensing device 10 and the motion analysis device 20 is increased and the cost is reduced. Can do.

DESCRIPTION OF SYMBOLS 10 Sensing apparatus 101 Sensing means 102 Data transmission means 20 Motion analysis apparatus 201 Data reception means 202 Data storage means 203 Data processing means 2031, 2131 First frequency conversion means 2032, 2132 Logarithmic conversion means 2033 Frequency inverse conversion means 2034, 2134 Low frequency Pass filter 2035, 2135 Second frequency conversion means 2036 Pace detection means 2133 First frequency reverse conversion means 2136 Exponential conversion means 2137 Second frequency reverse conversion means 2138 Motion determination means 2139 Teacher data storage means 204 Display means

Claims (7)

Movement detection means for detecting body movement;
First time / frequency converting means for generating first frequency data by converting first time data indicating time-series motion detected by the motion detecting means into frequency domain data;
Nonlinear conversion means for nonlinearly converting the first frequency data generated by the first time / frequency conversion means to generate second frequency data;
First frequency / time conversion means for generating second time data by converting the second frequency data generated by the nonlinear conversion means into time domain data;
Low frequency component extraction means for extracting low frequency components from the second time data generated by the first frequency / time conversion means to generate third time data;
Second time / frequency converting means for generating third frequency data by converting the third time data generated by the low frequency component extracting means into frequency domain data;
Pace detecting means for detecting the pace of the movement from a peak position appearing in the third frequency data generated by the second time / frequency converting means;
A motion detection device comprising: The first time / frequency conversion means converts the first time data cut out at a predetermined sampling period and a predetermined cut-out interval from the detection results of the motion detection means into first frequency data. Generate frequency data for
The low frequency component extraction unit extracts a low frequency component from second time data using a predetermined threshold corresponding to the predetermined sampling period and the predetermined cut-out interval. The motion detection device described. The pace detecting means interpolates the third frequency data which is a discrete data string by a predetermined interpolation method, and detects the pace of the movement from a peak position appearing in the third frequency data after the interpolation. The motion detection device according to claim 2, wherein: Non-linear inverse transformation means for producing fourth frequency data by performing inverse transformation of nonlinear transformation by the nonlinear transformation means on the third frequency data generated by the second time / frequency conversion means;
Second frequency / time conversion means for converting the fourth frequency data generated by the non-linear inverse conversion means into time domain data to generate fourth time data;
Data storage means for storing the fourth time data generated in advance for each type of exercise;
The fourth time data generated by the second frequency / time conversion means based on the motion detected by the motion detection means is compared with the fourth time data stored in the data storage means. , Exercise type determination means for determining the type of exercise corresponding to the movement detected by the movement detection means;
The motion detection device according to claim 1, further comprising: The exercise type determination means includes
A pattern between the fourth time data generated by the second frequency / time conversion unit based on the motion detected by the motion detection unit and the fourth time data stored in the data storage unit Similarity calculation means for calculating similarity;
Of the fourth time data stored in the data storage means, fourth time data having a high pattern similarity calculated by the similarity calculation means is extracted, and an exercise corresponding to the extracted fourth time data Determination result output means for outputting the type of
The motion detection device according to claim 4, comprising: A motion detection step for detecting body movement;
A first time-frequency conversion step of generating first frequency data by converting first time data indicating time-series motion detected in the motion detection step into data in the frequency domain;
A non-linear conversion step of generating non-linear conversion of the first frequency data generated in the first time / frequency conversion step to generate second frequency data;
A first frequency / time conversion step of generating second time data by converting the second frequency data generated in the nonlinear conversion step into time domain data;
A low frequency component extracting step of extracting a low frequency component from the second time data generated in the first frequency / time conversion step to generate third time data;
A second time / frequency conversion step of generating third frequency data by converting the third time data generated in the low frequency component extraction step into frequency domain data;
A pace detecting step of detecting a pace of the movement from a peak position appearing in the third frequency data generated in the second time-frequency conversion step;
The motion detection method characterized by including. A motion detection function that detects the movement of the body,
A first time-frequency conversion function for generating first frequency data by converting first time data indicating time-series motion detected by the motion detection function into data in the frequency domain;
A non-linear conversion function for generating non-linear conversion of the first frequency data generated by the first time / frequency conversion function to generate second frequency data;
A first frequency / time conversion function for generating second time data by converting the second frequency data generated by the nonlinear conversion function into time domain data;
A low frequency component extraction function for extracting low frequency components from the second time data generated by the first frequency / time conversion function to generate third time data;
A second time-frequency conversion function for generating third frequency data by converting the third time data generated by the low-frequency component extraction function into data in the frequency domain;
A pace detecting function for detecting the pace of the movement from a peak position appearing in the third frequency data generated by the second time / frequency converting function;
A program to make a computer realize.

Images