CN105912142A - Step recording and behavior identification method based on acceleration sensor - Google Patents

Step recording and behavior identification method based on acceleration sensor Download PDF

Info

Publication number
CN105912142A
CN105912142A CN201610082327.7A CN201610082327A CN105912142A CN 105912142 A CN105912142 A CN 105912142A CN 201610082327 A CN201610082327 A CN 201610082327A CN 105912142 A CN105912142 A CN 105912142A
Authority
CN
China
Prior art keywords
step
acceleration
waveform
movement
steps
Prior art date
Application number
CN201610082327.7A
Other languages
Chinese (zh)
Other versions
CN105912142B (en
Inventor
黄伟
李建
Original Assignee
深圳市爱康伟达智能医疗科技有限公司
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 深圳市爱康伟达智能医疗科技有限公司 filed Critical 深圳市爱康伟达智能医疗科技有限公司
Priority to CN201610082327.7A priority Critical patent/CN105912142B/en
Priority claimed from CN201610082327.7A external-priority patent/CN105912142B/en
Publication of CN105912142A publication Critical patent/CN105912142A/en
Application granted granted Critical
Publication of CN105912142B publication Critical patent/CN105912142B/en

Links

Classifications

    • GPHYSICS
    • G01MEASURING; TESTING
    • G01CMEASURING DISTANCES, LEVELS OR BEARINGS; SURVEYING; NAVIGATION; GYROSCOPIC INSTRUMENTS; PHOTOGRAMMETRY OR VIDEOGRAMMETRY
    • G01C22/00Measuring distance traversed on the ground by vehicles, persons, animals, or other moving solid bodies, e.g. using odometers, using pedometers
    • G01C22/006Pedometers
    • AHUMAN NECESSITIES
    • A43FOOTWEAR
    • A43BCHARACTERISTIC FEATURES OF FOOTWEAR; PARTS OF FOOTWEAR
    • A43B3/00Footwear characterised by the shape or the use
    • A43B3/0005Footwear provided with electrical or electronic systems
    • AHUMAN NECESSITIES
    • A61MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61BDIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B5/00Detecting, measuring or recording for diagnostic purposes; Identification of persons
    • A61B5/103Detecting, measuring or recording devices for testing the shape, pattern, colour, size or movement of the body or parts thereof, for diagnostic purposes
    • A61B5/11Measuring movement of the entire body or parts thereof, e.g. head or hand tremor, mobility of a limb
    • A61B5/1123Discriminating type of movement, e.g. walking or running
    • AHUMAN NECESSITIES
    • A61MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61BDIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B5/00Detecting, measuring or recording for diagnostic purposes; Identification of persons
    • A61B5/68Arrangements of detecting, measuring or recording means, e.g. sensors, in relation to patient
    • A61B5/6801Arrangements of detecting, measuring or recording means, e.g. sensors, in relation to patient specially adapted to be attached to or worn on the body surface
    • A61B5/6802Sensor mounted on worn items
    • A61B5/6804Garments; Clothes
    • A61B5/6807Footwear
    • AHUMAN NECESSITIES
    • A61MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61BDIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B5/00Detecting, measuring or recording for diagnostic purposes; Identification of persons
    • A61B5/72Signal processing specially adapted for physiological signals or for diagnostic purposes
    • A61B5/7235Details of waveform analysis
    • A61B5/725Details of waveform analysis using specific filters therefor, e.g. Kalman or adaptive filters
    • GPHYSICS
    • G06COMPUTING; CALCULATING; COUNTING
    • G06FELECTRIC DIGITAL DATA PROCESSING
    • G06F3/00Input arrangements for transferring data to be processed into a form capable of being handled by the computer; Output arrangements for transferring data from processing unit to output unit, e.g. interface arrangements
    • G06F3/01Input arrangements or combined input and output arrangements for interaction between user and computer
    • G06F3/03Arrangements for converting the position or the displacement of a member into a coded form
    • G06F3/033Pointing devices displaced or positioned by the user, e.g. mice, trackballs, pens or joysticks; Accessories therefor
    • G06F3/0346Pointing devices displaced or positioned by the user, e.g. mice, trackballs, pens or joysticks; Accessories therefor with detection of the device orientation or free movement in a 3D space, e.g. 3D mice, 6-DOF [six degrees of freedom] pointers using gyroscopes, accelerometers or tilt-sensors
    • AHUMAN NECESSITIES
    • A61MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61BDIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B2560/00Constructional details of operational features of apparatus; Accessories for medical measuring apparatus
    • A61B2560/02Operational features
    • A61B2560/029Operational features adapted for auto-initiation
    • AHUMAN NECESSITIES
    • A61MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61BDIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B2562/00Details of sensors; Constructional details of sensor housings or probes; Accessories for sensors
    • A61B2562/02Details of sensors specially adapted for in-vivo measurements
    • A61B2562/0219Inertial sensors, e.g. accelerometers, gyroscopes, tilt switches
    • AHUMAN NECESSITIES
    • A61MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61BDIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B2562/00Details of sensors; Constructional details of sensor housings or probes; Accessories for sensors
    • A61B2562/02Details of sensors specially adapted for in-vivo measurements
    • A61B2562/0247Pressure sensors

Abstract

The invention discloses a step recording and behavior identification method based on an acceleration sensor. The method is characterized in that the placing method of the sensor is that the forward direction of a foot is the positive direction of an X-axis, the leftward direction is the positive direction of a Y-axis, and a foot lifting direction is the negative direction of a Z-axis. Since when a human body moves, the displacement on the X-axis direction is relatively large, and acceleration changes are relatively obvious and have obvious periodicity, motion step number can be identified and motion behaviors can be identified just by setting rational threshold values. By smoothing filtering and Kalman filtering, waveform is smoother and errors are reduced, so a system can accurately record steps in real time. The method can accurately distinguish various motions of a person. The method can calculate step number of various motions in real time. In addition, the method does not have very high computing capacity on a system, and a mobile phone or a common single-chip microcomputer (MCU) in common configuration can complete operation.

Description

一种基于加速传感器的记步与行为识别方法 Step one kind of note recognition and behavior based on an acceleration sensor

技术领域 FIELD

[0001]本发明属于一种基于加速传感器的记步与行为识别方法。 [0001] The present invention belongs to the pedometer and behavior recognition method based on the acceleration sensor.

背景技术 Background technique

[0002]早期的运动识别主要是基于视觉方式的,给定一段图像序列或者一个视频片段,识别出人物的运动类型。 [0002] The Early motion is based on visual recognition mode, a given section of a sequence of images or video clip, the motion type of the character recognition. 基于视觉的方法具有交互自然,提取的特征信息丰富等优点,但该方法在实际应用中也有一些局限性,需要克服很多问题。 Feature-based visual interactive natural method to extract information rich, etc., but the method in practical applications, there are some limitations to overcome a lot of problems. 如环境中的光照条件,人物在摄像机前的位置,场地的大小等。 The environment light, the person in front of the camera position, the size of the venue and the like. 传感器具有价格便宜,携带方便,不受场地限制等优点,随着这些设备的发展,运动识别又被带入了一片新的研究领域,补充了传统基于视觉的运动识别方法在实际应用中的不足,促使了运动识别在日常生活中的应用。 Sensor is cheap, easy to carry, subject to site constraints, etc., with the development of these devices, motion recognition has been brought into a new area of ​​research, supplement the shortcomings of traditional identification methods based on visual motion in the practical application of to promote the application of motion recognition in daily life. 这一技术已经被用在行为障碍病人的康复状况监视,老年人突发疾病预防监视等应用中。 This technology has been used in the rehabilitation of patients with behavioral disorders to monitor the situation, the elderly sudden illness prevention surveillance applications. 常用的传感器有加速度传感器,陀螺仪,麦克风等,一些内置传感器的设备如Apple iPhone,Nintendo Wiimote等,这些无线设备的发展使得大范围的交互应用成为可能,如智能家庭,混合现实等应用。 Commonly used sensors acceleration sensor, a gyroscope, a microphone, etc., some of the built-in sensor devices such as Apple iPhone, Nintendo Wiimote, etc., which makes the development of wireless devices interact with a wide range of possible applications, such as smart home, mixed reality applications.

[0003]对于使用加速度传感器进行运动识别而言,主要问题有三:一为如何快速自动地分割传感器输出的加速度信号,以达到在线的进行运动分割的目的,为后续的在线识别做准备;二为如何建立有效的分类模型,以达到高效准确的对运动进行分类识别的目的;三为如何采用适当的方法,在运动结束之间进行识别,提高交互感。 [0003] using an acceleration sensor for motion recognition, the main problems are three: one for the acceleration signal how quickly and automatically divides the sensor output, in order to achieve line motion segmentation to prepare for online recognition subsequent; two for building an effective model for classification, in order to achieve efficient and accurate classification of the motion; third how appropriate method, between the end of the movement identification, a sense of more interactive. 本发明将以这三个问题为基本出发点,对运动识别过程中的关键问题进行分析,解决以上提到的主要技术问题,实现一个高效的在线运动识别系统。 The present invention will be three problems as the basic starting point, the movement of the key issues identified during the analysis, mainly to solve the above mentioned technical problems, to achieve an efficient line motion recognition system.

[0004]对于加速度信号分割问题,很多研究工作都是将传感器信号手动分割好,作为训练和测试的数据库。 [0004] For the acceleration signal segmentation, a lot of research work are the sensor signal manual segmentation well as training and testing database. 这样降低了信号处理的负担,并且数据比较理想化,在此基础上排除了数据的影响,可以对比分析识别算法的性能。 This reduces the burden of signal processing, and more idealistic data, on the basis of the negative influence of data can be compared to analyze the performance of the recognition algorithm. 但是实际应用中,手动的方法交互感不好,不便于操作和应用,因此我们需要对信号进行在线的分割处理;对于分类模型的选取,现阶段大多数研究与相应的系统采用动态时间卷曲算法(DTW)和隐马尔科夫模型方法(HMM) ,DTff算法所需的训练数据较少,并且能够动态的更新匹配的模板。 However, the practical application, the manual method is not good sense of interaction, it is not easy to operate and use, so we need the signal line division processing; classification model selected for this stage most of the research and the corresponding system uses dynamic time-warped Algorithm (DTW) and hidden Markov model method (HMM), DTff algorithm requires less training data, and can be dynamically updated to match the template. 但该算法的运算速度会随着待识别的时序数据的长度以及模板的数量的增大而明显的减慢,HMM方法用一个状态表示当前动作,但是很多全身性动作比较复杂,无法仅仅用一个状态充分表示出来,因此需要两个或多个状态变量来表示,本发明采用Fused HMM方法,解决了单独的一个HMM无法对具有相关关系的两个时序序列同时进行建模的问题,对于具有交互过程的全身性动作具有很好的描述能力,并且当一个HMM信息丢失时另一个HMM仍能正常工作,增加了算法的鲁棒性;对于提前进行运动识别问题,当前主要的处理方法是当一个运动完成之后再去调用识别过程,在有些应用中这种延迟感会降低用户体验度。 However, the operation speed of the algorithm will increase as the number of sequential data length to be recognized and significantly slow down the template, the current operation of the HMM method represented by a state, but many more complex systemic action, not only with a state fully represented, requiring two or more state variables to represent the present invention employs Fused HMM method, an HMM alone solves the problem of not having a correlation between two time sequences while modeling for interactive systemic action process has a good ability to describe, and the other can still work HMM HMM when a message is lost, increasing the robustness of the algorithm; to advance the movement to identify the problem, the current main approach is when a after the completion of the movement again call identification process, in some applications, this sense of delay will reduce the user experience. 本发明采用了自回归的预测模型,利用已知帧数据,预测出未知的数据,通过对预测得到的数据进行分析,可以在运动结束之前即开始识别的过程,并达到提前识别的效果。 The present invention uses a prediction of the autoregressive model, using known frame data, predicted data of the unknown, by analyzing the prediction data obtained, i.e., may start before the end of the movement of the identification process, and to achieve early recognition.

发明内容 SUMMARY

[0005]本发明所要解决的技术问题是提供一种基于加速传感器的记步与行为识别方法。 [0005] The present invention solves the technical problem is to provide a pedometer and behavior recognition method based on an acceleration sensor.

[0006]本发明解决上述技术问题所采取的技术方案如下: [0006] The present invention solves the above technical problem adopted technical solution is as follows:

一种基于加速传感器的记步与行为识别方法,其特征在于,本方法所采用的传感器的放置方式为:脚向前的方向为X轴正方向,向左的方向为Y轴的正方向,抬脚方向为Z轴的负方向,同时,由于人体运动的时候X轴方向位移比较大,加速度变化也比较明显而且有很明显周期性;因此,只要设定合理的阈值就可以识别出运动的步数; Step one kind of note recognition and behavior based on an acceleration sensor, wherein the position sensor of the present method is employed: foot forward direction is the positive X-axis direction, the leftward direction is the positive direction of the Y axis, heels direction is negative Z-axis direction, while, due to the body movement when displaced in the X-axis direction is relatively large, the acceleration change is more obvious and has obvious periodicity; Therefore, if the set a reasonable threshold value can be identified motion Step count;

通过采集大量数据样本发现当X轴的加速度人体在运动时其加速度一定会大于一个阈值Ax,当加速度从小于Ax变到Ax以上,然后再由Ax以上变到Ax—下刚好对应人体抬脚和落脚动作,即识别出人体运动了一步,由于受到传感器存在一些误差,可能出现在一步内出现多个点的加速度在Ax附近徘徊,通过相应的方式计算就会出现多计算步数的情况,为了排除这种情况,根据人体最大的运动速度推算,人在一秒钟运动的步数不会超过5步,设传感器的采样率为25Hz,那么在25个采样点内计算的步数不能多余5步,由于传感器只放在一只鞋子内,人体运动5步时,其实一只脚最多运动了3步,所以I秒内的计算出的步数不能超过3步。 By collecting a large number of data samples found that when the X-axis acceleration of the body in which motion acceleration will be greater than a threshold value Ax, when the acceleration Ax becomes small in the above Ax, Ax and then from the above just changed to correspond to the human heels and Ax- settled operation, i.e., the body-movement identification step, the sensor due to the presence of some errors may occur in one step occurs in the vicinity of the plurality of points acceleration Ax hovering, calculated in a corresponding manner by calculating the number of steps of multi occurs, for excluded that, according to human estimated maximum velocity, the number of people in a second step of movement no more than five steps, the sensor sampling rate provided 25Hz, then calculated the number of steps in the sampling point 25 is not superfluous 5 step, since the sensor only be placed in a shoe, when human movement 5 steps, in fact, one foot up to the 3-step movement, so the number of steps calculated within seconds, I can not be more than 3 steps. 因此可以推算出人运动一步至少要大于8有个采样点,根据这个规则去除那些因误差而多计算出的步数,从而精确计算步数,具体步骤如下: Thus the movement of a person can be calculated step there must be at least 8 sample points, the number of steps that remove error due to multiple calculated based on this rule, to accurately calculate the number of steps, the following steps:

而精确计算步数,具体步骤如下: And accurate calculation of the number of steps, the following steps:

步骤一:通过智能鞋把采集到的加速度传感器数据实时发送给手机或者传送给通用单片机(MCU); Step one: transmitting real time collected by the smart shoe acceleration sensor data to the phone or transmitted to a general purpose microcontrollers (the MCU);

步骤二:将采集到的数据进行平滑滤波和卡尔曼滤波,使波形更加光滑减少误差; 步骤三:对平滑滤波后的数据进行分析计算出运动的步数; Step two: the collected data smoothing filtering and Kalman filtering, the waveform smoother to reduce errors; Step Three: smoothing the data is analyzed to calculate the number of steps of movement;

步骤四:对卡尔曼滤波后的数据切分出每一步的波形,分析波形的特征值,确认出人体运动状态; Step Four: Kalman filter data after each step Parsing waveform, waveform analysis of eigenvalues, confirmed human movement state;

步骤五:通过对两种数据的融合即可分析出人体各种运动状态的步数; Step Five: by fusion of two types of data to analyze the number of steps of the various human motion state;

通过对人体运动时加速度波形的进行分析,可以看出每一种运动的波形都存在着相应的周期性,而且在一个周期内不同运动的波形是不一样的,我们对波形的特征值加以区分就可以区分出每一种运动。 By analysis of the body movement when the acceleration waveform, each waveform can be seen that there is a corresponding movement of both periodic, but in a different motion cycle of the waveform is not the same, we distinguish between waveform feature values We can distinguish every movement.

[0007]优选的,所述的基于加速传感器的记步与行为识别方法的步骤一中的智能鞋为一种可以采集人体运动过程中的加速度信息,并实时通过蓝牙发送给手机的鞋子或者传送给通用单片机(MCU)。 [0007] Preferably, the step referred to based on the behavior recognition step and an acceleration sensor in the intelligent shoe as a body motion acceleration information can be collected in the process, or transmitted in real time through the shoe to a Bluetooth mobile phone for general purpose microcontrollers (MCU).

[0008]进一步地,优选的,所述的基于加速传感器的记步与行为识别方法的步骤三中,在手机端获取到X,Y,Z轴的加速度数值后,把原始数据复制为两份,一份通过平滑滤波,一份通过卡尔曼滤波的方式去消除误差。 After [0008] Further preferred, based on the behavior recognition odograph acceleration sensor of the three-step, the mobile terminal acquires acceleration values ​​X, Y, Z-axis, the original data is copied two , through a smoothing filter, a Kalman filter to eliminate errors manner.

[0009]进一步地,优选的,所述的基于加速传感器的记步与行为识别方法的步骤四中,平滑滤波采用简单平均法进行,为求邻近像元点的平均亮度值,经过平滑滤波后的数据用于计算运动的步数。 [0009] Further, preferably, in the step referred to based on the behavior of the acceleration sensor recognition step four, smoothing filtering is a simple average method, the average luminance value for the sake of pixels adjacent to the point, after the smoothing the data used to calculate the number of steps of movement.

[0010]进一步地,优选的,所述的基于加速传感器的记步与行为识别方法的步骤五中卡尔曼滤波后的数据显示每一种运动的人体不同的加速度值在一定程度上反应了运动的剧烈程度,因此可以用加速度的大小来区分步行,快走和跑步,合加速度的计算公式如下: 其中,a为合加速度,ax,ay,az分别为传感器测出的X轴,Y轴,Z轴的加速度,求出一个周期内合加速度的平均值a',根据a'的大小即可区分出走路,快走和跑步;区分出走和跑之后,在此基础上进一步分析,提取出波形的特征值,根据特征值对波形进行分类,即可确认人体的运动状态; [0010] Further, preferably, the acceleration sensor based on the step of the step counter behavior recognition methods fifth Kalman filtered data for each display a different body motion acceleration value to a certain extent reflects the movement the severity, it is possible to use the magnitude of acceleration to distinguish between walking, fast walking and running, calculated resultant acceleration is as follows: wherein, a is the resultant acceleration, ax, ay, az are sensors measure the X axis, Y axis, Z axis acceleration, the acceleration is obtained in a combined cycle average value a ', according to a' size to distinguish between a walk, brisk walking and running; and ran away after distinguishing, based on further analysis, the extracted characteristic waveform value, classified according to the waveform feature values, to confirm the state of motion of the body;

有关于特征值提取,计算出一个周期内波形的平均值,平均差,四分位差,离散系数,偏态系数等作为波形的特征值; About feature extraction, the calculated average value of the waveform in one cycle, the mean difference, quartile deviation, coefficient of variation, skewness coefficient as a characteristic value of the waveform;

通过对实际运动采样统计确定合理的阈值,即可精确区分出各种运动。 By statistical sampling to determine a reasonable actual movement threshold can accurately distinguish the various sports.

[0011]本发明采取了上述方案以后,借助于平滑滤波和卡尔曼滤波,使波形更加光滑减少误差,使得系统能够实时准确地记步;同时,还能够准确地区分出人的各种运动;其次,能够实时计算出各种各种运动的步数;再次,对系统的计算能要求不是很高,普通配置的手机或者通用单片机(MCU)即可完成运算。 [0011] The present invention takes the above-described embodiment after, by means of Kalman filtering and smoothing, reduce errors more smooth waveform, so that accurate real-time system can be referred to step; Meanwhile, it is possible to accurately separate the various movements of people; Secondly, the number of steps can be calculated in real time of various kinds of sports; again, the computing system can not very high, an ordinary mobile phone or a general configuration of a microcontroller (MCU) to complete the operation.

[0012]本发明的其它特征和优点将在随后的说明书中阐述,并且,部分地从说明书中变得显而易见,或者通过实施本发明而了解。 [0012] Other features and advantages of the invention will be set forth in the description which follows, and in part will be obvious from the description, or learned by practice of the present invention. 本发明的目的和其他优点可通过在所写的说明书、权利要求书、以及附图中所特别指出的结构来实现和获得。 The objectives and other advantages of the invention may be realized and attained by the written description, claims, and drawings structure particularly pointed out.

附图说明 BRIEF DESCRIPTION

[0013]下面结合附图对本发明进行详细的描述,以使得本发明的上述优点更加明确。 DRAWINGS The invention is described in detail [0013] below with reference to the above-described advantages of the present invention is such that more clearly. 其中, among them,

图1是本发明基于加速传感器的记步与行为识别方法的运动过程中X轴的加速度示意图; Figure 1 is a schematic view of the present invention is based on the acceleration movement the acceleration sensor and the step counter behavior identification process in the X-axis;

图2是本发明基于加速传感器的记步与行为识别方法的卡尔曼滤波后步行时加速度的波形示意图; FIG 2 is a waveform diagram of the present invention, the acceleration during walking steps after acceleration sensor referred Kalman filter is based on recognition and behavior;

图3是本发明基于加速传感器的记步与行为识别方法的卡尔曼滤波之后前脚掌着地跑步时加速度的波形示意图;; FIG 3 is a waveform of the acceleration of the present invention is based upon the former sole after running the Kalman filter acceleration sensor and the step counter behavior recognition methods schematic ;;

图4是本发明基于加速传感器的记步与行为识别方法的卡尔曼滤波之后上楼梯时加速度的波形示意图; FIG 4 is a waveform diagram of the present invention on the stairs after the acceleration step the acceleration sensor referred Kalman and behavior recognition method based filtering;

图5是本发明基于加速传感器的记步与行为识别方法的卡尔曼滤波之后下楼梯时加速度的波形示意图; FIG 5 is a schematic view of the present invention is based on the waveform after the acceleration sensor is noted in step Kalman behavior recognition method and the filtered acceleration down the stairs;

图6是本发明基于加速传感器的记步与行为识别方法的卡尔曼滤波之后全脚掌着地跑步时加速度的波形示意图; FIG 6 is a waveform diagram of the present invention, the acceleration when the whole foot after running the Kalman mind step acceleration sensor and the behavior recognition method based filtering;

图7是本发明基于加速传感器的记步与行为识别方法的卡尔曼滤波后后脚跟着地跑步时加速度的波形示意图; FIG 7 is a waveform diagram of the present invention, when the rear legs strike the acceleration after running the Kalman filter acceleration sensor and the step counter behavior recognition methods based;

图8是本发明基于加速传感器的记步与行为识别方法的卡尔曼滤波后快走时加速度的波形示意图; FIG 8 is a waveform diagram of the present invention, after the acceleration when the acceleration sensor is noted in step Kalman filter and behavior recognition method based brisk walking;

图9是本发明基于加速传感器的记步与行为识别方法的流程图。 FIG 9 is based on the recognition and behavior noted in step a flowchart of an acceleration sensor of the present invention.

具体实施方式 Detailed ways

[0014]以下将结合附图1-9及实施例来详细说明本发明的实施方式,借此对本发明如何应用技术手段来解决技术问题,并达成技术效果的实现过程能充分理解并据以实施。 [0014] conjunction with the accompanying drawings 1-9 and will be described in detail in Example embodiments of the present invention, how the present invention is applied whereby the technical means to solve the technical problem, and achieve the technical effect of the process can be fully understood and implemented according to embodiments . 需要说明的是,只要不构成冲突,本发明中的各个实施例以及各实施例中的各个特征可以相互结合,所形成的技术方案均在本发明的保护范围之内。 Incidentally, they do not constitute a conflict, various embodiments of the present invention and the various embodiments of the various features may be combined with each other, are within the scope of the technical solutions of the present invention are formed.

[0015] —种基于加速传感器的记步与行为识别方法,当手机端获取到X,Y,Z轴的加速度数值后,由于采样率,测量噪声等会对传感器的数据有一定的影响,导致数据误差很大,需要对原始数据进行滤波,本文采用把原始数据复制为两份,一份通过平滑滤波,一份通过卡尔曼滤波的方式去消除误差。 [0015] - Step kind referred to and behavior recognition methods based on an acceleration sensor, when the mobile terminal acquires acceleration values ​​X, Y, Z-axis, since the data have a sampling rate of the sensor, the measurement noise and the like to a certain extent, resulting in data error is large, it is necessary to filter the raw data, copy paper, the raw data into two, through a smoothing filter, a Kalman filter to eliminate errors manner.

[0016]空间域的平滑滤波一般采用简单平均法进行,就是求邻近像元点的平均亮度值。 Smoothing [0016] General spatial domain using a simple average method, it is to compute the average luminance value of the pixel adjacent to the point. 邻域的大小与平滑的效果直接相关,邻域越大平滑的效果越好,但邻域过大,平滑会使边缘信息损失的越大,从而使输出的图像变得模糊,而且平滑滤波会导致波形有一定的滞后性,不能实时反映出人体运动姿态。 The size of the smoothing effect is directly related to the neighborhood, the larger the better the effect of smoothing the neighborhood, the neighborhood is too large, the greater the edge information loss causes smooth, so that the output image becomes blurred, and the smoothing filter will lead waveform has a certain lag, it does not reflect the attitude of human motion in real time. 但是他却能很好的区分出人体运动的步数,经过平滑滤波的数据可以用来计算运动的步数。 However, he was able to distinguish the number of steps good body movement, after the smoothed data can be used to calculate the number of steps of movement.

[0017]本算法所采用的传感器的放置方式为:脚向前的方向为X轴正方向,向左的方向为Y轴的正方向,抬脚方向为Z轴的负方向。 [0017] The position sensor according to the algorithm used is: foot forward direction is the positive X-axis direction, the leftward direction is the positive Y-axis direction, a direction heels negative direction of the Z axis. 人体运动的时候X轴方向位移比较大,加速度变化也比较明显而且有很明显周期性(如图1所示)。 When the X-axis direction body movement displacement relatively large acceleration changes are more obvious and there was a clear periodicity (Figure 1). 只要设定合理的阈值就可以识别出运动的步数。 By setting a reasonable threshold value can identify the number of steps of movement.

[0018]通过采集大量数据样本发现当X轴的加速度人体在运动时其加速度一定会大于一个阈值(设为Ax),当加速度从小于Ax变到Ax以上,然后再由Ax以上变到Ax—下刚好对应人体抬脚和落脚动作,即识别出人体运动了一步,由于受到传感器存在一些误差,可能出现在一步内出现多个点的加速度在Ax附近徘徊,通过上面的方式计算就会出现多计算步数的情况,为了排除这种情况,根据人体最大的运动速度推算,人在一秒钟运动的步数不会超过5步,设传感器的采样率为25Hz,那么在25个采样点内计算的步数不能多余5步,由于传感器只放在一只鞋子内,人体运动5步时,其实一只脚最多运动了3步,所以I秒内的计算出的步数不能超过3步。 [0018] By collecting a large number of data samples found that when the X-axis acceleration of the body in which motion acceleration will be greater than a threshold value (set Ax), from less than when the acceleration Ax Ax changed to the above, and then change over to the Ax Ax- just under the corresponding human heels and settled operation, i.e., the body-movement identification step, the sensor due to some errors, the acceleration may occur in the presence of a plurality of points near hovering step Ax, calculated by the above manner will be multiple in the case of calculating the number of steps, in order to exclude this case, the maximum body velocity estimation, the number of people in a second step motion of no more than five steps, 25Hz, then the sampling point 25 disposed in the sensor sampling rate the number of steps calculated in step 5 is not superfluous, since the sensor only be placed in a shoe, when human movement 5 steps, in fact, one foot up to the 3-step movement, so the number of steps calculated within seconds, I can not be more than 3 steps. 因此可以推算出人运动一步至少要大于8有个采样点,根据这个规则去除那些因误差而多计算出的步数,从而达到精确计算步数的目的。 Thus the movement of a person can be calculated step there must be at least 8 sample points, the number of steps that remove error due to multiple calculated based on this rule, so as to achieve accurate calculation of the number of steps.

[0019]卡尔曼滤波通过系统输入输出观测数据,对系统状态进行最优估计的算法即保证了波形的信息,又使波形很平滑,给波形的特征值提取提供了方便。 [0019] Kalman filter observation data input and output through the system, the system state estimation algorithm is to ensure optimal information waveform, and the waveform is smoothed, to extract the value of the waveform feature provides convenience. 为根据波形特征值区分运动状态提供了可能。 It provides the possibility to distinguish between the movement state according to the value of the waveform feature. 经过卡尔曼滤波之后的波形如下图所示: After waveform after the Kalman filter as shown below:

通过对以上几种常见运动时加速度波形的进行分析,可以看出每一种运动的波形都存在着周期性,而且在一个周期内不同运动的波形是不一样的,我们对波形的特征值加以区分就可以区分出每一种运动。 By analysis of the acceleration waveform when several more common movement, each waveform can be seen that there are cyclical movement, but in a different motion cycle of the waveform is not the same, we feature value of the waveform to be distinction can distinguish every movement.

[0020]加速度的值在一定程度上反应了运动的剧烈程度,因此可以用加速度的大小来区分步行,快走和跑步。 Value [0020] of the acceleration in a certain extent reflects the intensity of exercise, so the magnitude of acceleration can be used to distinguish between walking, fast walking and running. 合加速度的计算公司如下: Company resultant acceleration is calculated as follows:

a=V(a_x'2+a_y'2+a_z'2 ) a = V (a_x'2 + a_y'2 + a_z'2)

a:合加速度,a_x,a_y,a_z分别为传感器测出的X轴,Y轴,Z轴的加速度求出一个周期内合加速度的平均值a',根据a'的大小即可区分出走路,快走和跑步。 a: acceleration together, a_x, a_y, a_z were measured X-axis sensor, acceleration in the Y-axis, Z-axis average value of acceleration within a combined cycle a ', according to a' size to distinguish between a walk, brisk walking and jogging. 区分出走和跑之后,在此基础上进一步分析,提取出波形的特征值,根据特征值对波形进行分类,即可确认人体的运动状态。 And ran away after distinguishing, based on further analysis, extracts a feature value of the waveform, the waveform classification according to the characteristic value, to confirm the state of motion of the human body. 有关于特征值提取,计算出一个周期内波形的平均值,平均差,四分位差,离散系数,偏态系数等作为波形的特征值。 About feature extraction, feature values ​​calculated average value, mean difference, quartile deviation, coefficient of variation, skewness coefficient in one cycle of the waveform as the waveform. 通过对实际运动采样统计确定合理的阈值,即可精确区分出各种运动。 By statistical sampling to determine a reasonable actual movement threshold can accurately distinguish the various sports.

[0021]本发明采取了上述方案以后,借助于平滑滤波和卡尔曼滤波,使波形更加光滑减少误差,使得系统能够实时准确地记步;同时,还能够准确地区分出人的各种运动;其次,能够实时计算出各种各种运动的步数;再次,对系统的计算能要求不是很高,普通配置的手机或者通用单片机(MCU)即可完成运算。 [0021] The present invention takes the above-described embodiment after, by means of Kalman filtering and smoothing, reduce errors more smooth waveform, so that accurate real-time system can be referred to step; Meanwhile, it is possible to accurately separate the various movements of people; Secondly, the number of steps can be calculated in real time of various kinds of sports; again, the computing system can not very high, an ordinary mobile phone or a general configuration of a microcontroller (MCU) to complete the operation.

Claims (5)

1.一种基于加速传感器的记步与行为识别方法,其特征在于,本方法所采用的传感器的放置方式为:脚向前的方向为X轴正方向,向左的方向为Y轴的正方向,抬脚方向为Z轴的负方向,同时,由于人体运动的时候X轴方向位移比较大,加速度变化也比较明显而且有很明显周期性;因此,只要设定合理的阈值就可以识别出运动的步数; 通过采集大量数据样本发现当X轴的加速度人体在运动时其加速度一定会大于一个阈值Ax,当加速度从小于Ax变到Ax以上,然后再由Ax以上变到Ax—下刚好对应人体抬脚和落脚动作,即识别出人体运动了一步,由于受到传感器存在一些误差,可能出现在一步内出现多个点的加速度在Ax附近徘徊,通过相应的方式计算就会出现多计算步数的情况,为了排除这种情况,根据人体最大的运动速度推算,人在一秒钟运动的步数不会超过5步, A pedometer and behavior recognition method based on an acceleration sensor, wherein the position sensor of the present method is employed: foot forward direction is the positive X-axis direction, leftward direction is a Y axis positive direction, heels direction is negative Z-axis direction, while, due to the body movement when displaced in the X-axis direction is relatively large, the acceleration change is more obvious and has obvious periodicity; Therefore, if the set a reasonable threshold value can be identified number of steps moving; by collecting a large number of data samples found that when the X-axis acceleration of the body in motion its acceleration will be greater than a threshold value Ax, when the acceleration from less than Ax changed to Ax above, then the Ax of the above variations to the Ax- just corresponding to human heels and settled operation, i.e., the body-movement identification step, the sensor due to the presence of some errors may occur in one step occurs in the vicinity of the plurality of points acceleration Ax hovering calculated corresponding way there will be a multi-step calculation case number, in order to exclude this case, the body's largest movement speed projections, the number of people in a second step the movement of no more than five steps, 设传感器的采样率为25Hz,那么在25个采样点内计算的步数不能多余5步,由于传感器只放在一只鞋子内,人体运动5步时,其实一只脚最多运动了3步,所以I秒内的计算出的步数不能超过3步,因此可以推算出人运动一步至少要大于8有个采样点,根据这个规则去除那些因误差而多计算出的步数,从而精确计算步数,具体步骤如下: 步骤一:通过智能鞋把采集到的加速度实时发送给手机或者传送给通用单片机(MCU); 步骤二:将采集到的数据进行平滑滤波和卡尔曼滤波,使波形更加光滑减少误差; 步骤三:对平滑滤波后的数据进行分析计算出运动的步数; 步骤四:对卡尔曼滤波后的数据切分出每一步的波形,分析波形的特征值,确认出人体运动状态; 步骤五:通过对两种数据的融合即可分析出人体各种运动状态的步数; 通过对人体运动时加速度波形的进行分析, Set sensor sampling rate of 25Hz, then calculate the number of steps in the 25 sampling points can not be superfluous step 5, since the sensor only be placed in a shoe, when human movement 5 steps, in fact, one foot up to the 3-step movement, Therefore, the number of steps computed in the I sec no more than 3 steps, it is possible to calculate at least the person moves step is greater than 8 have sampling points, removing those few steps due to an error and multiple calculated based on this rule, so that accurate calculation step number, the following steps: step one: to collect the acceleration transmitted to the real-time through the smart phone or to a general shoe microcontroller (the MCU); step 2: the data collected Kalman filtering and smoothing filter, the waveform smoother reduce errors; step three: smoothing the data is analyzed to calculate the number of steps of movement; step four: data separated after each step of the Kalman filter cut waveform characteristic value of the waveform analysis, it was confirmed that human movement state ; step five: by fusion of two types of data to analyze the number of steps of the various motion states of the human body; by analyzing human movement during acceleration waveform, 以看出每一种运动的波形都存在着相应的周期性,而且在一个周期内不同运动的波形是不一样的,我们对波形的特征值加以区分就可以区分出每一种运动。 Seen that the waveform of each motion are present with respective periodic, but in a different motion cycle of the waveform is not the same, we distinguish between waveform feature values ​​can distinguish every movement.
2.根据权利要求1所述的基于加速传感器的记步与行为识别方法,其特征在于,所述步骤一中的智能鞋为一种可以采集人体运动过程中的加速度信息,并实时通过蓝牙发送给手机的鞋子,或者在通用单片机(MCU)上处理数据。 The step counter and behavior recognition method based on an acceleration sensor, wherein said 1, a step in the intelligent shoe as a body motion acceleration information can be collected in the process, and real-time transmission via Bluetooth claim to the phone's shoes, or processing data on the general purpose microcontroller (MCU).
3.根据权利要求1所述的基于加速传感器的记步与行为识别方法,其特征在于,所述步骤三中,在处理器获取到X,Y,Z轴的加速度数值后,把原始数据复制为两份,一份通过平滑滤波,一份通过卡尔曼滤波的方式去消除误差。 Step 3. Record and behavior recognition method based on an acceleration sensor, wherein one of the preceding claims, in step three, acquired X, Y, Z axis acceleration value in a processor, the raw data replication is two, by a smoothing filter, a Kalman filter to eliminate errors manner.
4.根据权利要求1所述的基于加速传感器的记步与行为识别方法,其特征在于,所述步骤四中,平滑滤波采用简单平均法进行,为求邻近像元点的平均亮度值,经过平滑滤波后的数据用于计算运动的步数。 Step 4. Write and behavior recognition method based on an acceleration sensor, wherein according to claim 1, in step four, smoothing filtering is a simple average method, the average luminance value for the sake of pixels adjacent to the point, after data smoothing filter for calculating the number of steps of movement.
5.根据权利要求1所述的基于加速传感器的记步与行为识别方法,其特征在于,所述步骤五中卡尔曼滤波后的数据显示每一种运动的人体不同的加速度值在一定程度上反应了运动的剧烈程度,因此可以用加速度的大小来区分步行,快走和跑步,合加速度的计算公式如下: 其中,a为合加速度,ax,ay,az分别为传感器测出的X轴,Y轴,Z轴的加速度,求出一个周期内合加速度的平均值a',根据a'的大小即可区分出走路,快走和跑步;区分出走和跑之后,在此基础上进一步分析,提取出波形的特征值,根据特征值对波形进行分类,即可确认人体的运动状态; 有关于特征值提取,计算出一个周期内波形的平均值,平均差,四分位差,离散系数,偏态系数等作为波形的特征值; 通过对实际运动采样统计确定合理的阈值,即可精确区分出各种运动。 According to claim pedometer and behavior recognition method based on an acceleration sensor, wherein said 1, the data show different Kalman filtering step 5 of each body motion acceleration value to a certain extent the reaction of the intensity of exercise, it is possible to use the magnitude of acceleration to distinguish between walking, fast walking and running, calculated resultant acceleration is as follows: wherein, a is the resultant acceleration, ax, ay, az are sensor measured X-axis, Y acceleration axis, Z-axis, within a determined period of engagement of the average acceleration a ', according to a' size to distinguish between a walk, brisk walking and running; and ran away after distinguishing, based on further analysis, extracted characteristic value of the waveform based on the characteristic value of the waveform classification, to confirm the state of motion of the human body; about feature extraction, the calculated average value of a cycle of the waveform, mean difference, quartile deviation, coefficient of variation, skewness coefficient as the characteristic value of the waveform; by determining the threshold value for a reasonable statistical sampling actual movement, can accurately distinguish the various sports.
CN201610082327.7A 2016-02-05 A kind of note step and Activity recognition method based on acceleration sensor CN105912142B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201610082327.7A CN105912142B (en) 2016-02-05 A kind of note step and Activity recognition method based on acceleration sensor

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN201610082327.7A CN105912142B (en) 2016-02-05 A kind of note step and Activity recognition method based on acceleration sensor
US15/339,848 US20170241797A1 (en) 2016-02-01 2016-10-31 Pedometer with Accelerometer and Foot Motion Distinguishing Method

Publications (2)

Publication Number Publication Date
CN105912142A true CN105912142A (en) 2016-08-31
CN105912142B CN105912142B (en) 2019-06-11

Family

ID=

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN106562508A (en) * 2016-10-19 2017-04-19 泉州迪特工业产品设计有限公司 Intelligent shoes for child and realization method therefor
CN106723612A (en) * 2016-11-21 2017-05-31 歌尔股份有限公司 Step counting system

Citations (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JP2008262522A (en) * 2007-04-11 2008-10-30 Aichi Micro Intelligent Corp Pedometer
CN102278998A (en) * 2010-03-25 2011-12-14 精工电子有限公司 Electronic equipment and procedures
CN102297701A (en) * 2010-06-22 2011-12-28 雅马哈株式会社 Pedometer
JP5176047B2 (en) * 2012-04-09 2013-04-03 アイチ・マイクロ・インテリジェント株式会社 Pedometer
CN103727954A (en) * 2013-12-27 2014-04-16 北京超思电子技术股份有限公司 Pedometer
US20140188431A1 (en) * 2012-11-01 2014-07-03 Hti Ip, Llc Method and system for determining whether steps have occurred
US20160001131A1 (en) * 2014-07-03 2016-01-07 Katarzyna Radecka Accurate Step Counting Pedometer for Children, Adults and Elderly

Patent Citations (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JP2008262522A (en) * 2007-04-11 2008-10-30 Aichi Micro Intelligent Corp Pedometer
CN102278998A (en) * 2010-03-25 2011-12-14 精工电子有限公司 Electronic equipment and procedures
CN102297701A (en) * 2010-06-22 2011-12-28 雅马哈株式会社 Pedometer
JP5176047B2 (en) * 2012-04-09 2013-04-03 アイチ・マイクロ・インテリジェント株式会社 Pedometer
US20140188431A1 (en) * 2012-11-01 2014-07-03 Hti Ip, Llc Method and system for determining whether steps have occurred
CN103727954A (en) * 2013-12-27 2014-04-16 北京超思电子技术股份有限公司 Pedometer
US20160001131A1 (en) * 2014-07-03 2016-01-07 Katarzyna Radecka Accurate Step Counting Pedometer for Children, Adults and Elderly

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN106562508A (en) * 2016-10-19 2017-04-19 泉州迪特工业产品设计有限公司 Intelligent shoes for child and realization method therefor
CN106723612A (en) * 2016-11-21 2017-05-31 歌尔股份有限公司 Step counting system

Also Published As

Publication number Publication date
US20170241797A1 (en) 2017-08-24

Similar Documents

Publication Publication Date Title
Bobick et al. The recognition of human movement using temporal templates
CN104995581B (en) Management of electronic equipment gesture detection
KR101208232B1 (en) Gesture recognize device, control method of recognize device and contol program
CN101344816B (en) Human-machine interaction method and device based on sight tracing and gesture discriminating
Chen et al. A survey of depth and inertial sensor fusion for human action recognition
US9075453B2 (en) Human eye controlled computer mouse interface
CN101763515B (en) Real-time gesture interaction method based on computer vision
CN102789568B (en) Gesture identification method based on depth information
CN101393599B (en) Based on facial expression control method of game characters
JP5845002B2 (en) An image processing apparatus and method, and program
Bujari et al. Movement pattern recognition through smartphone's accelerometer
Planinc et al. Introducing the use of depth data for fall detection
Yu et al. An online one class support vector machine-based person-specific fall detection system for monitoring an elderly individual in a room environment
JP2006523878A (en) Method and system for determining the target pose from the image
CN101719015B (en) Method for positioning finger tips of directed gestures
CN101038671A (en) Tracking method of three-dimensional finger motion locus based on stereo vision
US20160086023A1 (en) Apparatus and method for controlling presentation of information toward human object
CN103353935B (en) 3d dynamic gesture recognition method for an intelligent home system
US8781991B2 (en) Emotion recognition apparatus and method
CN102567502B (en) Supplemented with biometric identification device identification
JP6025845B2 (en) Object attitude retrieval apparatus and method
Li et al. Automatic recognition of sign language subwords based on portable accelerometer and EMG sensors
Kusakunniran et al. Gait recognition across various walking speeds using higher order shape configuration based on a differential composition model
CN102663364A (en) Imitated 3D gesture recognition system and method
CN102200834B (en) Television control-oriented finger-mouse interaction method

Legal Events

Date Code Title Description
C06 Publication
C10 Entry into substantive examination