JP2013022188A - Gait analyzing method, gait analyzer, and program of the same - Google Patents

Abstract

PROBLEM TO BE SOLVED: To accurately analyze the gait of a user.SOLUTION: A gait analyzer 1 includes: an acceleration information reception part 101 for time-sequentially receiving first acceleration data measured when a user walks; a smoothing processing part 102 for performing smoothing processing to the first acceleration data; a walking cycle calculation part 103 for performing frequency analysis to each of second acceleration data after the smoothing processing and the first acceleration data, extracting the candidate of a peak frequency from the power spectrum of the second acceleration data, detecting the peak frequency near the candidate from the power spectrum of the first acceleration data, and calculating the walking cycle T of the user by the peak frequency; an autocorrelation calculation part 104 for calculating the autocorrelation function of the first acceleration data; and a peak detection part 105 for detecting a peak in the autocorrelation function on the basis of the walking cycle T.

Description

  The present invention relates to a gait analysis method, a gait analysis device, and a program thereof.

  In the autocorrelation function for the acceleration data measured when the user walks, the odd-numbered peak indicates the time when the left and right legs are walked, and the even-numbered peak compares the objectivity of the right legs or the left legs. Therefore, in research on gait analysis, the first peak and the second peak shown in FIG. 6 are used as gait parameters.

  However, the time-series change in acceleration during walking varies greatly depending on the scene, depending on the geographical factors at the user's walking position, the user's health condition, and the like. Therefore, the noise added to the acceleration data varies depending on the scene. Therefore, when detecting an autocorrelation peak in an autocorrelation function calculated using acceleration data, it is necessary to consider the influence of these noises that vary depending on the scene.

Non-Patent Document 1 discloses a technique for detecting a peak by performing a three-point comparison on an audio signal that has been smoothed using a moving average method in order to detect a peak in a sound signal. .
Further, Non-Patent Document 2 discloses a technique for extracting a position where a sign is positive from negative using smoothing differentiation, and extracting a position that is equal to or greater than a threshold from the extracted positions as a peak.

Shinya Kotosaka, Stefan Schaal, “Rhythmic motion generation of robots using neural oscillators”, Journal of the Robotics Society of Japan, Vol.19, No.1, pp.116-123, 2001 Masataka Goto, “Real-time music scene description system: rust section detection method”, Information Processing Society of Japan Information and Music Society Research Report, Vol.2002, No.100, pp.27-34, 2002

FIG. 6 shows, as an example, an autocorrelation function of acceleration during walking of a rehabilitation patient. FIG. 7 shows the result of smoothing the autocorrelation function shown in FIG. In FIG. 7, peaks detected by the three-point comparison are indicated by circles.
Since the technique of Non-Patent Document 1 is a method of calculating a peak from a smoothed signal waveform, as shown in FIG. 7, a portion other than the autocorrelation peak in the vicinity of the gait period that is important for gait analysis is a peak. Will be extracted.

  Further, in the technique of Non-Patent Document 2, an unnecessary peak is deleted by adding a threshold process. However, since the peak size may differ depending on the user's walking scene, an appropriate threshold value is set. It becomes difficult.

  The present invention has been made paying attention to the above circumstances, and an object of the present invention is to provide a gait analysis method, a gait analysis device, and a program thereof that enable a user's gait to be analyzed with high accuracy. is there.

  To achieve the above object, according to a first aspect of the present invention, there is provided a gait analysis method, a gait analysis device and a program thereof for analyzing a user's gait, wherein the first acceleration measured during the user's walking Receiving data in a time series, performing a smoothing process on the first acceleration data, performing frequency analysis on each of the second acceleration data and the first acceleration data after the smoothing process, and 2 extracting a peak frequency candidate from the power spectrum of the acceleration data, detecting a peak frequency in the vicinity of the candidate from the power spectrum of the first acceleration data, calculating the walking period T of the user from the peak frequency, An autocorrelation function of the first acceleration data is calculated, and a peak in the autocorrelation function is detected based on the walking cycle T.

  According to the first aspect, it is possible to reduce the influence of high frequency component noise included in the measured acceleration data on the peak detection of the autocorrelation function. This makes it possible to accurately detect peaks in autocorrelation functions that are useful for gait analysis, even for acceleration data during walking that is difficult to detect with only smoothing processing such as a low-pass filter or threshold processing. It becomes.

According to a second aspect of the present invention, in the first aspect, the peak is detected by detecting, as the autocorrelation first peak, the maximum autocorrelation coefficient in the delay time near the walking cycle T in the autocorrelation function. To do.
According to the second aspect, the first peak in the autocorrelation function useful for gait analysis is prevented by preventing erroneous detection as a peak of the high frequency component of the power spectrum caused by noise included in the measured acceleration data. Can be detected with high accuracy.

According to a third aspect of the present invention, in the second aspect, the peak is detected in the autocorrelation function when the nth autocorrelation peak (n is a natural number) in the autocorrelation function is detected. The maximum autocorrelation coefficient in the delay time in the vicinity of the walking cycle T × n is detected as the nth autocorrelation peak.
According to the third aspect, the n-th peak used as a gait parameter in research in gait analysis can be obtained accurately and easily.

  That is, according to the present invention, it is possible to provide a gait analysis method, a gait analysis device, and a program thereof that enable a user's gait to be analyzed with high accuracy.

The figure which shows the structure of the gait analyzer which concerns on one Embodiment of this invention. The block diagram which shows the function structure of the gait analyzer of FIG. The flowchart which shows operation | movement of the gait analyzer of FIG. The figure which shows an example of the power spectrum of acceleration data. The figure which shows the calculation result of the autocorrelation function of acceleration data. The figure which shows the calculation result of the autocorrelation function of general acceleration data. The figure which shows the smoothing process result of the autocorrelation function of FIG.

Hereinafter, a gait analysis method, a gait analysis device, and a program thereof according to an embodiment of the present invention will be described with reference to the drawings.
FIG. 1 is a diagram showing the configuration of a gait analyzer according to an embodiment of the present invention.
The gait analyzer 1 is configured by a computer having a CPU (Central Processing Unit), a memory, and the like, and receives acceleration data obtained by an acceleration sensor 2 carried by the user. As an embodiment, a method in which the acceleration sensor 2 transmits acceleration data to the gait analysis device 1 regardless of whether it is wired or wireless, or the gait analysis device 1 as a calculation processing terminal integrated with the acceleration sensor 2 is possible. It is also possible to use a method of configuring

  FIG. 2 is a block diagram showing a functional configuration of the gait analyzer 1. The gait analyzing apparatus 1 includes an acceleration information receiving unit 101, a smoothing processing unit 102, a walking cycle calculation unit 103, an autocorrelation calculation unit 104, and a peak detection unit 105. These units are realized by the CPU of the gait analyzer 1 and a control program executed on the memory.

  The acceleration information receiving unit 101 includes a communication interface or the like, and receives acceleration data measured by the acceleration sensor 2 when the user walks in time series by wire or wireless. Further, the acceleration information receiving unit 101 may be configured to read acceleration data stored in a recording medium such as a memory card.

The smoothing processing unit 102 performs a smoothing process on the acceleration data received by the acceleration information receiving unit 101.
The walking cycle calculation unit 103 receives the acceleration data (first acceleration data) received by the acceleration information receiving unit 101 and the smoothed acceleration data (second acceleration data) output from the smoothing processing unit 102, respectively. A user's walking cycle is calculated from a power spectrum obtained by frequency analysis by a method described later.

The autocorrelation calculating unit 104 calculates an autocorrelation function of the acceleration data received by the acceleration information receiving unit 101.
The peak detection unit 105 detects a peak in the autocorrelation function calculated by the autocorrelation calculation unit 104 based on the walking cycle calculated by the walking cycle calculation unit 103.

Next, the operation of the gait analyzer configured as described above will be described in detail.
The acceleration information receiving unit 101 receives acceleration data of a predetermined time length from the acceleration sensor 2 at predetermined time intervals. Alternatively, acceleration data is received every time an event based on a predetermined rule occurs. Hereinafter, the time series data of the acceleration data from the received time t1 to time tn is defined as _{at1: tn} . The acceleration information receiving unit 101 transmits the received acceleration data _{at1: tn} to the smoothing processing unit 102 and the walking cycle calculation unit 103.

The smoothing processing unit 102 performs a smoothing process on the acceleration data _{at1: tn} received from the acceleration information receiving unit 101. Thereby, the influence which the noise of the high frequency component of acceleration data _{at1: tn} has on the peak detection of the autocorrelation function can be reduced. As an example of the smoothing method, there is a smoothing process using a low-pass filter. This can be realized by multiplying the acceleration data _{at1: tn} by a first-order lag transfer function G (s) shown in the following equation (1).

ただし、式（１）において、Ｋ，ω_ｃ，ｓはそれぞれ、ローパスフィルタの通過域での利得、遮断周波数、ラプラス変換の変数とする。これ以下、平滑化された加速度データを平滑化データａ´_{ｔ１：ｔｎ}と呼ぶ。平滑化処理部１０２は、平滑化データａ´_{ｔ１：ｔｎ}を歩行周期算出部１０３に送信する。

However, in Equation (1), K, ω _c , and s are variables for the gain, cutoff frequency, and Laplace transform in the pass band of the low-pass filter, respectively. Hereinafter, the smoothed acceleration data is referred to as smoothed data a ′ _{t1: tn} . The smoothing processing unit 102 transmits the smoothed data a ′ _{t1: tn} to the walking cycle calculation unit 103.

Walking period calculating unit 103, the acceleration data _a received from the acceleration information receiving unit 101 _{t1: tn} and smoothed data received from the smoothing processing unit 102 _a't1: frequency analysis using FFT or the like for each of _tn Is performed to calculate the user's walking cycle. Hereinafter, the details of the walking cycle calculation process will be described with reference to the flowchart of FIG.

In step S1 of FIG. 3, the walking period calculating unit 103, the acceleration data _{a t1} from the acceleration information receiving unit _101: receiving a _tn, smoothing processing unit 102 smoothes data from _a't1: receiving a _tn.
In step S2, the walking period calculating unit 103, the received acceleration data _{a t1: tn} and smoothed data _a't1: in frequency space by performing a discrete Fourier transform by using a FFT for each of _tn Calculate the power spectrum component. FIG. 4A shows an example of a result obtained by _performing a discrete Fourier transform on the acceleration data a _{t1: tn} , and FIG. 4B shows an example of a result obtained by _performing a discrete Fourier transform on the smoothed data a ′ _{t1: tn} . Show.

In step S3, the walking period calculating section 103, smoothed data _a'in the frequency space calculated in step S2 _t1: power spectrum components _f't1 of _tn: of _tn, the detection of a frequency having a maximum power spectrum Do. As a detection method, for example, a local peak is sequentially searched by comparing three points before and after the power spectrum in ascending order from the lowest frequency excluding the DC component, and then the maximum power is found from the searched peak group. detecting a peak with a spectrum as the peak frequency candidate _f'p.

In step S4, the walking period calculating unit 103 uses the peak frequency candidate _f'p detected in step S3, the acceleration data received in the step S1 _{a t1:} power spectrum component information _f of _{tn t1:} the _tn detecting a peak frequency f _p from within. Specifically, for the power spectrum component f _{t1: tn} of the acceleration data a _{t1: tn} , the peak frequency f _p having the maximum power spectrum peak at a frequency near the peak frequency candidate f ′ _p is detected. As a peak detection method, three points before and after the power spectrum with respect to the frequency axis are compared in the vicinity of the peak frequency candidate f ′ _p . As can be seen from FIG. 4 (b), this way, by detecting the peak frequency f _p, acceleration data a _t1: from erroneous detection as a peak of the high frequency component of the power spectrum caused by noise contained in _tn You can expect to prevent it.

In step S5, the walking period calculating unit 103 calculates a walking period T of the user by the following equation (2) from the peak frequency f _p detected in step S4.

ステップＳ６において、歩行周期算出部１０３は、上記ステップＳ５で算出したユーザの歩行周期Ｔを自己相関算出部１０４及びピーク検出部１０５に送信し、歩行周期算出処理を終了する。

In step S6, the walking cycle calculation unit 103 transmits the user's walking cycle T calculated in step S5 to the autocorrelation calculation unit 104 and the peak detection unit 105, and ends the walking cycle calculation process.

The autocorrelation calculation unit 104 calculates an autocorrelation function of the acceleration data _{at1: tn} . The autocorrelation function represents a time-series change of the autocorrelation coefficient at each delay time with the vertical axis representing the autocorrelation coefficient R (k) and the horizontal axis representing the delay time k. An example of the calculation result of the autocorrelation function is shown in FIG. Further, the autocorrelation calculation unit 104 transmits the calculated autocorrelation function to the peak detection unit 105. For example, the autocorrelation coefficient R (k) of the acceleration data _{at1: tn} is calculated by the following equation (3).

なお、式（３）において、ａ（ｔ）は、時刻ｔにおける加速度データを表す。また、ｋは、ｋサンプル分の遅れ時間を表し、ｎは、歩容解析に用いる加速度データの総サンプル数を表す。

In equation (3), a (t) represents acceleration data at time t. K represents a delay time corresponding to k samples, and n represents the total number of acceleration data samples used for gait analysis.

The peak detection unit 105 detects a peak (autocorrelation peak) in the autocorrelation function received from the autocorrelation calculation unit 104 using the user's walking cycle T received from the walking cycle calculation unit 103. Specifically, an autocorrelation coefficient having the maximum value in the delay time in the vicinity of the walking cycle T is searched, and this value is set as the autocorrelation first peak _Rp1 .

As a search method, the maximum value of the autocorrelation coefficient is searched for by comparing three points before and after the autocorrelation function in the vicinity of the walking cycle T. Similarly, when detecting the nth autocorrelation peak (n is a natural number), the autocorrelation coefficient having the maximum value in the delay time in the vicinity of the walking cycle n × T is detected. The autocorrelation n-th peak R _{pn is} assumed. FIG. 5 shows a detection example from the first peak to the fourth peak.

  As described above, according to the present embodiment, paying attention to the fact that the delay time at the autocorrelation peak represents the walking period, in the autocorrelation function, the maximum delay time in the vicinity of the user's walking period extracted by frequency analysis. The autocorrelation coefficient is taken as the autocorrelation peak. Thus, by taking into account the characteristics of walking, autocorrelation peaks in the vicinity of the walking cycle that are useful in gait analysis can be accurately detected even for autocorrelation functions whose waveforms are disturbed as shown in FIG. Is possible.

  Note that the present invention is not limited to the above-described embodiment as it is, and can be embodied by modifying the constituent elements without departing from the scope of the invention in the implementation stage. Further, various inventions can be formed by appropriately combining a plurality of constituent elements disclosed in the embodiment. For example, some components may be deleted from all the components shown in the embodiment. Furthermore, you may combine suitably the component covering different embodiment.

  DESCRIPTION OF SYMBOLS 1 ... Gait analyzer 2 ... Acceleration sensor 101 ... Acceleration information receiving part 102 ... Smoothing process part 103 ... Walking period calculation part 104 ... Autocorrelation calculation part 105 ... Peak detection part

Claims (7)

A method of analyzing a user's gait by a computer,
A receiving step of receiving, in time series, first acceleration data measured when the user walks;
A smoothing step for performing a smoothing process on the first acceleration data;
Frequency analysis is performed on each of the second acceleration data and the first acceleration data after the smoothing process, a candidate of a peak frequency is extracted from the power spectrum of the second acceleration data, and the power of the first acceleration data Detecting a peak frequency in the vicinity of the candidate from the spectrum, and calculating a walking period T of the user based on the peak frequency;
An autocorrelation calculating step of calculating an autocorrelation function of the first acceleration data;
A gait analysis method comprising: a peak detection step of detecting a peak in the autocorrelation function based on the walking period T.   The gait analysis method according to claim 1, wherein the peak detecting step detects a maximum autocorrelation coefficient in a delay time in the vicinity of the walking cycle T in the autocorrelation function as an autocorrelation first peak. .   In the peak detection step, when detecting the autocorrelation n-th peak (n is a natural number) in the autocorrelation function, the maximum autocorrelation coefficient in the delay time near the walking cycle T × n in the autocorrelation function is calculated. The gait analysis method according to claim 2, wherein the gait analysis method is detected as an autocorrelation nth peak. An apparatus for analyzing a user's gait,
Receiving means for receiving, in time series, first acceleration data measured when the user walks;
Smoothing means for performing a smoothing process on the first acceleration data;
Frequency analysis is performed on each of the second acceleration data and the first acceleration data after the smoothing process, a candidate of a peak frequency is extracted from the power spectrum of the second acceleration data, and the power of the first acceleration data A walking cycle calculating means for detecting a peak frequency in the vicinity of the candidate from a spectrum and calculating the walking cycle T of the user based on the peak frequency;
Autocorrelation calculating means for calculating an autocorrelation function of the first acceleration data;
A gait analyzing apparatus comprising: peak detecting means for detecting a peak in the autocorrelation function based on the walking period T.   The gait analyzer according to claim 4, wherein the peak detecting means detects a maximum autocorrelation coefficient in a delay time in the vicinity of the walking cycle T in the autocorrelation function as an autocorrelation first peak. .   When detecting the autocorrelation n-th peak (n is a natural number) in the autocorrelation function, the peak detecting means calculates a maximum autocorrelation coefficient in a delay time in the vicinity of the walking cycle T × n in the autocorrelation function. 6. The gait analyzer according to claim 5, wherein the gait analyzer is detected as an autocorrelation nth peak.   A gait analysis program for causing a computer to function as each means constituting the gait analysis device according to any one of claims 4 to 6.

Images