WO2021004510A1 - 基于传感器的分离式部署的人体行为识别健康管理*** - Google Patents

基于传感器的分离式部署的人体行为识别健康管理*** Download PDF

Info

Publication number
WO2021004510A1
WO2021004510A1 PCT/CN2020/101145 CN2020101145W WO2021004510A1 WO 2021004510 A1 WO2021004510 A1 WO 2021004510A1 CN 2020101145 W CN2020101145 W CN 2020101145W WO 2021004510 A1 WO2021004510 A1 WO 2021004510A1
Authority
WO
WIPO (PCT)
Prior art keywords
user
data
module
behavior
recognition
Prior art date
Application number
PCT/CN2020/101145
Other languages
English (en)
French (fr)
Inventor
许深颐
王若梅
周凡
林格
Original Assignee
中山大学
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 中山大学 filed Critical 中山大学
Publication of WO2021004510A1 publication Critical patent/WO2021004510A1/zh

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
    • G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V40/00—Recognition of biometric, human-related or animal-related patterns in image or video data
    • G06V40/20—Movements or behaviour, e.g. gesture recognition
    • G—PHYSICS
    • G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
    • G16H—HEALTHCARE INFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR THE HANDLING OR PROCESSING OF MEDICAL OR HEALTHCARE DATA
    • G16H20/00—ICT specially adapted for therapies or health-improving plans, e.g. for handling prescriptions, for steering therapy or for monitoring patient compliance
    • G16H20/30—ICT specially adapted for therapies or health-improving plans, e.g. for handling prescriptions, for steering therapy or for monitoring patient compliance relating to physical therapies or activities, e.g. physiotherapy, acupressure or exercising
    • G—PHYSICS
    • G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
    • G16H—HEALTHCARE INFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR THE HANDLING OR PROCESSING OF MEDICAL OR HEALTHCARE DATA
    • G16H50/00—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics
    • G16H50/30—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics for calculating health indices; for individual health risk assessment

Definitions

  • the invention relates to the field of pattern recognition of machine learning in human-computer interaction and artificial intelligence, and in particular to a human behavior recognition health management system based on separate deployment of sensors.
  • Human behavior recognition is an important issue of human-computer interaction in the field of pervasive computing. It plays an important role in promoting new types of human-computer interaction and making computers better understand and assist users in completing tasks.
  • the human behavior recognition problem is theoretically a pattern recognition problem in machine learning.
  • image and video recognition and sensor recognition Both solutions have many corresponding studies.
  • the identification scheme generally requires a fixed place, has low portability, and is not suitable for individual users.
  • traditional machine learning methods and deep learning methods target specific behaviors, and there are already better solutions in offline computing environments.
  • Portable computing devices such as smart phones or smart bracelets can provide a more flexible boundary carrier for the human behavior recognition system, and at the same time provide more exploration for the application of the human behavior recognition system, such as individual user sports and health monitoring. .
  • the purpose of the present invention is to overcome the shortcomings of the existing system and propose a separate deployment of a sensor-based human behavior recognition health management system.
  • the present invention proposes a separate deployment of sensor-based human behavior recognition health management system, the system includes:
  • the client includes a user interaction module and a data collection module, which are deployed on the user's personal terminal such as a smart phone or smart bracelet to interact with the user and collect user behavior data.
  • the server includes a model recognition module, a data analysis module, and a suggestion module, which are deployed on a remote host or server to identify the behavior data of the human body and perform corresponding data analysis to provide suggestions.
  • the client and the server communicate through the network.
  • the user interaction module is composed of user personal information management, personal behavior records, and suggestion reminders, and provides users with basic interactive operations, including user personal information management, displaying the history of their personal behaviors, and displaying suggestions.
  • the data collection module collects user behavior data, uses the sensor of the hardware where the client is located to collect corresponding data, including 3-axis acceleration sensor data, inertial sensor data, and preprocesses the collected data to reduce The amount of data transmitted over the network.
  • the model recognition module adopts a machine learning method and uses the computing hardware of the server to quickly and accurately recognize the data uploaded by the user, and feedback the recognition result to the user interaction module as feedback, and pass the recognition result to Suggested modules.
  • the data analysis module analyzes the user's movement in the past period of time according to the user's historical behavior records and the user's personal status and gives corresponding suggestions.
  • the suggestion module makes corresponding suggestions to the user according to the personal status set by the user, such as whether the amount of exercise reaches the standard, whether certain exercises should be reduced too much, and the sitting time is too long to get up and exercise.
  • the sensor-based separated deployment human behavior recognition health management system proposed in the present invention provides a separated human behavior recognition system that is easy to deploy, provides more comprehensive recognition behavior types and more complete and effective service reminders, and also The accuracy and speed of recognition are further improved, and it is convenient to provide users with more in-depth personalized services.
  • Figure 1 is a system architecture diagram of an embodiment of the present invention
  • Fig. 2 is a flowchart of a data collection module according to an embodiment of the present invention.
  • Fig. 1 is a system architecture diagram of an embodiment of the present invention. As shown in Fig. 1, the system includes:
  • the client and the server are composed of two parts.
  • the client includes a user interaction module and a data collection module, which are deployed on the user's personal terminal such as a smart phone or smart bracelet for interacting with the user and collecting user behavior data.
  • the server includes a model recognition module, a data analysis module, and a suggestion module, which are deployed on a remote host or server to identify the behavior data of the human body and perform corresponding data analysis to provide suggestions.
  • the client and the server communicate through the network.
  • the user interaction module is composed of user personal information management, personal behavior records, suggestion reminders, and provides users with basic interactive operations, including user personal information management, displaying the history of their personal behaviors, displaying suggestions, etc. .
  • the user uploads personal information in the client interactive module, while the collection module collects user behavior data and preprocesses it.
  • the personal information management sub-module can submit and modify the user's personal information, including basic information such as name, age, gender, height, weight, etc. This information will be synchronized to the personal database of the server, and the suggestion module of the server will be combined with the user Personal information and behavioral data provide corresponding suggestions.
  • Personal behavior record sub-module which can view daily behavior records in the form of charts, including the occurrence and duration of various behaviors, and the amount of exercise generated.
  • the suggestion reminder sub-module the suggestion message of the suggestion module of the server is sent back to the module and displayed to the user, while assisting some simple and basic reminder functions: sedentary reminder, when the sitting time exceeds 1 hour, it will raise activity reminder; Excessive exercise reminder, exercise energy consumption exceeds a certain range to remind rest and supplement energy.
  • FIG. S1-2, Figure 2 is a flowchart of a data acquisition module, which is composed of a sensor acquisition sub-module and a data pre-processing sub-module.
  • the data collection module collects the user's behavior data, uses the sensor of the hardware where the client is located to collect the corresponding data, including 3-axis acceleration sensor data, inertial sensor data, and preprocesses the collected data to reduce network transmission data the amount.
  • the three-axis gyroscope and the three-axis acceleration sensor collect data at a frequency of 20HZ.
  • the data is smoothed by sliding window filtering.
  • the data is simplified through operations.
  • the data is standardized through normalization operations.
  • the data is divided into paragraphs through sliding windows.
  • the sensor collection sub-module collects data by using the sensor of the client hardware.
  • a smart phone such as an Android phone
  • the phone’s 3-axis acceleration sensor and inertial sensor (gyro) are collected by the corresponding application program interface.
  • the frequency of collecting data is set to 20 Hz.
  • Si is the data at time i
  • the data preprocessing sub-module performs simple preprocessing operations on the data after collecting the data.
  • the data is smoothed by sliding window filtering, and the window size is 2 seconds.
  • the specific filtering formula is, Where w is the window size.
  • the specific synthesis formula is, Where As represents the composite value of acceleration or gyroscope data, and Ax, Ay, and Az represent the components of its x, y, and z axes, respectively.
  • Vmax and Vmin are the maximum and minimum values of the same feature. After normalization, the data is scaled to between 0 and 1.
  • the original data is segmented using sliding window segmentation.
  • the segmented window size is 2 seconds.
  • X min min ⁇ X 1 ,X 2 ,...,X w ⁇
  • the user interaction module synchronizes the information with the server and receives the information feedback from the server through the network information exchange interface.
  • the characteristic data finally obtained by the data acquisition module is sent to the server through the network information exchange interface of the client.
  • the model recognition module adopts a machine learning method, uses the computing hardware of the server to quickly and accurately recognize the data uploaded by the user, and feeds back the recognition result to the user interaction module as feedback, and transmits the recognition result Give suggestions for modules.
  • the model recognition module uses the user behavior data from the client to use the hardware calculation of the server to recognize the behavior through the integrated learning method in machine learning.
  • the server uses the xgboost model to train a behavior recognition model in an offline state, and uses the trained model to recognize user behavior data from the client. By replacing the new model on the server side, the recognized behavior types can be effectively expanded, making it more flexible to adapt to the user's customized behavior needs.
  • the behavior data passed by the client is segmented feature data.
  • the parameter settings for the xgboost model are as follows: objective is the training target parameter, select "multi:softmax” for multi-classification, and set the number of categories parameter num_class to the target category number 11; eval_metric is the evaluation index parameter, select "merror” to indicate Multi-class error rate; lambda and alpha are the shadows of L1 and L2 regular penalty items, the parameter is set to 0, eta is the learning step size, set to 0.3; max_depth is the maximum depth, set to 12.
  • the behaviors recognized by this system include typing and writing in a sitting state, walking, running, going upstairs, going downstairs, cycling, push-ups, sit-ups, squats, rope skipping, and a total of 11 behaviors, namely recognition behaviors.
  • a, a ⁇ writing, typing , walk, run, upstairs, downstairs, riding, pushup, situp, squat, ropeskipping ⁇ by xgboost model
  • user behavior data F i is identified as the behavior of the user is most likely carried out a i, A i ⁇ A.
  • the data analysis module analyzes the user's exercise situation in the past period of time according to the user's historical behavior record and the user's personal status and gives corresponding suggestions, and forwards the analyzed suggestions to the suggestion module.
  • This module will record the user's behavior data history, and according to the change history, statistics and analysis of the user's daily, weekly, and monthly various types of behaviors, combined with the user's height and weight and other information on different types of behavior Make suggestions to increase or decrease, such as for exercise behavior. If the amount of exercise is insufficient, it is recommended to increase the amount of exercise the next day; if the amount of exercise is large, it is recommended to rest the next day; if the running time is too long, it is recommended to reduce the amount of exercise , Protect the knees.
  • the suggestion module makes corresponding suggestions to the user according to the personal status set by the user, such as whether the amount of exercise reaches the standard, whether certain exercises should be reduced too much, and the sitting time is too long to get up and exercise.
  • the suggestion module returns the current behavior suggestions and the analyzed suggestions to the client, and at the same time gives corresponding suggestions based on real-time behavior monitoring, such as reminding the user based on the behavior time of the user's sitting state, reminding the user to get up and move, for different sitting Behavior, if writing for a long time, remind the active wrist, if the typing time is long, it will additionally remind the user to take a proper rest with the eyes, the user interaction module of the client will show the user and provide suggestion reminders.
  • the embodiment of the present invention proposes a separate deployment of a sensor-based human behavior recognition health management system, provides a separate human behavior recognition system that is easy to deploy, provides more comprehensive recognition behavior types and more complete and effective service reminders, and at the same time It also further improves the accuracy and speed of recognition, which is convenient to provide users with more in-depth personalized services.
  • the program can be stored in a computer-readable storage medium, and the storage medium can include: Read only memory (ROM, Read Only Memory), random access memory (RAM, Random Access Memory), magnetic disk or optical disk, etc.

Landscapes

  • Engineering & Computer Science (AREA)
  • Health & Medical Sciences (AREA)
  • General Health & Medical Sciences (AREA)
  • Data Mining & Analysis (AREA)
  • Medical Informatics (AREA)
  • Public Health (AREA)
  • Theoretical Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Physics & Mathematics (AREA)
  • Epidemiology (AREA)
  • Primary Health Care (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Artificial Intelligence (AREA)
  • Databases & Information Systems (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Biophysics (AREA)
  • Evolutionary Biology (AREA)
  • Evolutionary Computation (AREA)
  • Pathology (AREA)
  • General Engineering & Computer Science (AREA)
  • Physical Education & Sports Medicine (AREA)
  • Biomedical Technology (AREA)
  • Psychiatry (AREA)
  • Social Psychology (AREA)
  • Human Computer Interaction (AREA)
  • Multimedia (AREA)
  • Measurement Of The Respiration, Hearing Ability, Form, And Blood Characteristics Of Living Organisms (AREA)
  • Medical Treatment And Welfare Office Work (AREA)

Abstract

本发明公开了基于传感器的分离式部署的人体行为识别健康管理***。本发明***包括客户端的用户交互模块、数据采集模块，服务端的模型识别模块、数据分析模块、建议模块。客户端部署在智能手机或者智能手环等用户个人终端上，用于与用户交互及采集用户的行为数据。服务端部署在远程的主机或服务器上，用于识别人体的行为数据以及进行相应的数据分析，从而提供建议。客户端与服务端通过网络进行通信。本发明提供了便于部署的分离式的人体行为识别***，提供了更加全面的识别行为的种类以及更加完善有效的服务提醒，同时也进一步提高了识别的准确度以及速度，便于针对用户提供更加深度个性化的服务。

Description

基于传感器的分离式部署的人体行为识别健康管理*** 技术领域
本发明涉及人机交互、人工智能中机器学习的模式识别领域，具体涉及基于传感器的分离式部署的人体行为识别健康管理***。
背景技术
人体行为识别问题作为普适计算领域中人机交互问题的一个重要问题，对于推动新型人机交互方式以及使得计算机更好的理解和辅助用户完成任务起着重要的作用。人体行为识别问题从理论上来讲属于机器学习中的模式识别问题，目前主要有两种解决方案，以图像视频识别为主和以传感器识别为主，两种方案都有许多相应的研究，但图像识别方案一般需要固定的场所，便携性较低，不适合个人用户使用。对于解决这两种方案，传统机器学习方法以及深度学习方法针对特定的行为，在离线计算环境下已经有较好的解决方案。
随着计算设备的不断发展以及传感器技术的进步，移动计算设备开始展现出不可估量的潜力。智能手机或智能手环等便携式计算设备能够为人体行为识别***提供一个更加灵活边界的载体，同时也为人体行为识别***的应用提供了更多的探索，如个人用户的运动和健康监控等方面。
现有的各种解决方案的缺点在于：
使用简单的统计特征如波形特征学习，对不同用户具有偏差，需要每次针对不同用户重新学习，时间成本较高。对于采用单一机器学习的方法， 准确率不够高。而对于基于卷积神经网络的深度学习方法进行模式分类的方案，其训练好的分类模型较大，在移动端使用时计算速度较慢，处理的时间较长，不具有实时性。并且这些方案只使用了当前行为信息，没有利用到历史行为信息，也没有利用用户的个人信息，不能针对用户提供更加深度个性化的服务。并且目前智能手环对行为进行自动识别的种类还比较少，不能够全面的识别用户的行为，不能够根据行为种类提供准确的用户提醒信息。而手动识别不同类别则需要用户自己上传运动数据，过程繁琐。
发明内容
本发明的目的是克服现有***不足，提出了基于传感器的分离式部署的人体行为识别健康管理***。提供便于部署的分离式的人体行为识别***，能够提供更加全面的识别行为的种类以及更加完善有效的服务提醒，同时也进一步提高识别的准确度，以及更快识别速度。
为了解决上述问题，本发明提出了基于传感器的分离式部署的人体行为识别健康管理***，所述***包括：
由客户端与服务端两部分构成，客户端包括用户交互模块，数据采集模块，部署在智能手机或者智能手环等用户个人终端上，用于与用户进行交互和采集用户的行为数据。服务端包括模型识别模块，数据分析模块，建议模块，部署在远程的主机或服务器上，用于识别人体的行为数据以及进行相应的数据分析，从而提供建议。客户端与服务端通过网络进行通信。
优选地，所述用户交互模块由用户个人信息管理，个人行为记录，建议提醒构成，为用户提供基本的交互操作，包括用户个人信息管理，展示 其个人的行为的历史情况，展示建议提示等。
优选地，所述数据采集模块对用户的行为数据进行采集，使用客户端所在硬件的传感器收集相应的数据，包括3轴加速度传感器数据，惯性传感器数据，并将采集到的数据进行预处理从而降低网络传输的数据量。
优选地，所述模型识别模块采用机器学习的方法，利用服务端的计算硬件，将用户上传的数据进行快速准确的识别，并将识别的结果反馈给用户交互模块作为反馈，以及将识别结果传递给建议模块。
优选地，所述数据分析模块根据用户的历史行为记录以及用户的个人状况分析用户过去一段时间的运动情况以及给出相应的建议。
优选地，所述建议模块根据用户设置的个人状态对用户提出相应的建议，如运动量是否达到标准，某些运动是否过多应该减少，静坐时间过长应起身进行运动等。
本发明提出的基于传感器的分离式部署的人体行为识别健康管理***，提供了便于部署的分离式的人体行为识别***，提供了更加全面的识别行为的种类以及更加完善有效的服务提醒，同时也进一步提高了识别的准确度以及速度，便于针对用户提供更加深度个性化的服务。
附图说明
为了更清楚地说明本发明实施例或现有技术中的技术方案，下面将对实施例或现有技术描述中所需要使用的附图作简单地介绍，显而易见地，下面描述中的附图仅仅是本发明的一些实施例，对于本领域普通技术人员来讲，在不付出创造性劳动的前提下，还可以根据这些附图获得其它的附 图。
图1是本发明实施例的***架构图；
图2是本发明实施例的数据采集模块的流程图。
具体实施方式
下面将结合本发明实施例中的附图，对本发明实施例中的技术方案进行清楚、完整地描述，显然，所描述的实施例仅仅是本发明一部分实施例，而不是全部的实施例。基于本发明中的实施例，本领域普通技术人员在没有作出创造性劳动前提下所获得的所有其他实施例，都属于本发明保护的范围。
图1是本发明实施例的***架构图，如图1所示，该***包括：
S1，客户端与服务端两部分构成，客户端包括用户交互模块，数据采集模块，部署在智能手机或者智能手环等用户个人终端上，用于与用户进行交互和采集用户的行为数据。服务端包括模型识别模块，数据分析模块，建议模块，部署在远程的主机或服务器上，用于识别人体的行为数据以及进行相应的数据分析，从而提供建议。客户端与服务端通过网络进行通信。
S1具体如下：
S1-1，所述用户交互模块由用户个人信息管理，个人行为记录，建议提醒构成，为用户提供基本的交互操作，包括用户个人信息管理，展示其个人的行为的历史情况，展示建议提示等。用户在客户端交互模块上传个人信息，同时采集模块采集用户的行为数据并预处理。
(1)个人信息管理子模块，可以提交和修改用户的个人信息，包括姓 名，年龄，性别，身高，体重等基本信息，该信息会同步到服务端的个人数据库中，服务端的建议模块会结合用户个人信息和行为数据提供相应的建议。
(2)个人行为记录子模块，该子模块可以通过图表的形式查看每日行为记录，包括各种行为发生和持续的时间，产生的运动量。
(3)建议提醒子模块，服务端的建议模块的建议消息回传到该模块展示给用户，同时辅助一些简单基础的提醒功能：久坐提醒，监测到静坐时间超过1小时则会提出活动提醒；运动过量提醒，运动消耗能量超过一定范围提出休息和补充能量的提醒。
S1-2，图2是数据采集模块的流程图，所述数据采集模块由传感器采集子模块和数据预处理子模块构成。数据采集模块通过对用户的行为数据进行采集，使用客户端所在硬件的传感器收集相应的数据，包括3轴加速度传感器数据，惯性传感器数据，并将采集到的数据进行预处理从而降低网络传输的数据量。
三轴陀螺仪和三轴加速度传感器按照20HZ频率采集数据。
数据通过滑动窗口滤波平滑处理。
数据通过成操作简化。
数据通过归一化操作标准化。
数据通过滑动窗口分成段落。
对每一段数据计算其特征。
(1)传感器采集子模块，通过利用客户端硬件的传感器进行数据收集，在智能手机上，如安卓手机上，依靠对应的应用程序接口调用手机的3轴加 速度传感器，惯性传感器(陀螺仪)收集人体运动时的3轴加速度数据以及3轴陀螺仪的数据，在智能穿戴设备上，同样依靠对应的应用程序接口调用其3轴加速度传感器和3轴陀螺仪收集数据。采集数据的频率设置为20赫兹。
收集到的原始数据集为S，S＝[S 1,S 2,...,S t]表示1到t时刻的数据。其中Si为i时刻的数据，S i＝[Acc,Gyro]＝[Acc x,Acc y,Acc z,Gyro x,Gyro y,Gyro z]。
(2)数据预处理子模块在收集到数据后将数据进行简单的预处理操作。首先通过滑动窗口滤波的方式平滑数据，窗口大小取2秒。具体的滤波的公式为，
Figure PCTCN2020101145-appb-000001
其中w为窗口大小。
然后将3轴加速度和3轴陀螺仪数据合成进一步减少数据量，具体的合成的公式为，
Figure PCTCN2020101145-appb-000002
其中As表示加速度或陀螺仪数据的合成值，Ax，Ay，Az分别代表其x，y，z轴的分量。
对合成后的数据进行归一化操作，采用线性归一化的方法，归一化的公式为
Figure PCTCN2020101145-appb-000003
其中Vmax和Vmin为同一特征的最大值和最小值，经过归一化后数据被缩放到0到1之间。
最后使用滑动窗口分段的方式将原始数据进行分段，分段的窗口大小取2秒，分段的公式为Y i＝[X i,X i+1,...,X i+w-1]，其中Xi为第i个原始数据，w为分段窗口大小。
分段后提取每段的统计特征，包括最大值，最小值，均值，方差，偏度，峰度六种特征，提取统计特征的公式如下。
最大值：X max＝max{X 1,X 2,...,X w}
最小值：X min＝min{X 1,X 2,...,X w}
均值：
Figure PCTCN2020101145-appb-000004
方差：
Figure PCTCN2020101145-appb-000005
标准差：
Figure PCTCN2020101145-appb-000006
偏度：
Figure PCTCN2020101145-appb-000007
峰度：
Figure PCTCN2020101145-appb-000008
用户交互模块的与服务端的信息同步以及接收服务端的信息反馈通过网络信息交换的接口。数据采集模块最终得到的特征数据通过客户端的网络信息交换接口发送给服务端。
S1-3，所述模型识别模块采用机器学习的方法，利用服务端的计算硬件，将用户上传的数据进行快速准确的识别，并将识别的结果反馈给用户交互模块作为反馈，以及将识别结果传递给建议模块。
模型识别模块将从客户端传来的用户行为数据利用服务端的硬件计算，通过机器学习中的集成学习的方法进行行为识别。
服务端采用xgboost模型在离线状态下训练行为识别模型，并使用训练好的模型对客户端传来的用户行为数据进行识别。通过在服务端更换新的模型可以有效的扩充所识别的行为种类，使得其能够更加灵活的适应用户的定制行为需求。客户端传递过来的行为数据为分段特征数据，每一段F i为两个六元特征向量Facc和Fgyro，分别代表加速度数据向量和陀螺仪数据向量，F acc＝F gyro＝[max,min,avg,var,skew,kurt]。
对于xgboost模型的参数设置如下：objective为训练目标参数，选择”multi:softmax”进行多分类，同时要设置类别个数的参数num_class为目标类别数目11；eval_metric为评估指标参数，选择”merror”表示多分类错误率；lambda和alpha为L1和L2正则惩罚项影子，参数设为0，eta为学习步长，设置为0.3；max_depth为最大深度，设置为12。
本***所识别的行为有静坐状态的打字和书写，运动状态下的行走，跑步，上楼，下楼，骑自行车，俯卧撑，仰卧起坐，深蹲，跳绳，共计11种行为，即识别行为集合A，A＝{writing,typing,walk,run,upstairs,downstairs,riding,pushup,situp,squat,ropeskipping}通过xgboost模型，用户行为数据F i被识别为用户最可能的进行的行为A i,A i∈A。
S1-4，所述数据分析模块根据用户的历史行为记录以及用户的个人状况分析用户过去一段时间的运动情况以及给出相应的建议，将分析后的建议转发给建议模块。
该模块会记录用户的行为数据历史，并且根据改历史情况对用户的每日，每周，每月的各种类型的行为运动进行统计和分析，结合用户的身高体重等信息对不同类型的行为提出建议增加或减少，如针对运动行为的建议，若运动量不足则建议次日增加运动量；若俯卧撑，仰卧起坐等健身运动，运动量大则建议次日适当休息；若跑步时间过长则建议减少运动量、保护膝盖。
S1-5，所述建议模块根据用户设置的个人状态对用户提出相应的建议， 如运动量是否达到标准，某些运动是否过多应该减少，静坐时间过长应起身进行运动等。
建议模块将当前行为建议和分析后的建议回传给客户端，同时根据实时行为的监测给与相应的建议，如根据用户的静坐状态的行为时间进行提醒，提醒用户起身活动，针对不同的静坐行为，若长时间书写则提醒活动手腕，若打字时间较长则额外提醒用户用眼过度进行适当休息，由客户端的用户交互模块展示给用户并提供建议提醒。
本发明实施例提出基于传感器的分离式部署的人体行为识别健康管理***，提供了便于部署的分离式的人体行为识别***，提供了更加全面的识别行为的种类以及更加完善有效的服务提醒，同时也进一步提高了识别的准确度以及速度，便于针对用户提供更加深度个性化的服务。
本领域普通技术人员可以理解上述实施例的各种方法中的全部或部分步骤是可以通过程序来指令相关的硬件来完成，该程序可以存储于一计算机可读存储介质中，存储介质可以包括：只读存储器(ROM，Read Only Memory)、随机存取存储器(RAM，Random Access Memory)、磁盘或光盘等。
另外，以上对本发明实施例所提供的基于传感器的分离式部署的人体行为识别健康管理***进行了详细介绍，本文中应用了具体个例对本发明的原理及实施方式进行了阐述，以上实施例的说明只是用于帮助理解本发明的方法及其核心思想；同时，对于本领域的一般技术人员，依据本发明的思想，在具体实施方式及应用范围上均会有改变之处，综上所述，本说明书内容不应理解为对本发明的限制。

Claims (6)

  1. 基于传感器的分离式部署的人体行为识别健康管理***，其特征在于，所述***包括：
    由客户端与服务端两部分构成，客户端包括用户交互模块、数据采集模块，部署在智能手机或者智能手环等用户个人终端上，用于与用户交互和采集用户的行为数据。服务端包括模型识别模块、数据分析模块、建议模块，部署在远程的主机或服务器上，用于识别人体的行为数据以及进行相应的数据分析，从而提供建议。客户端与服务端通过网络进行通信。
  2. 如权利要求1所述的基于传感器的分离式部署的人体行为识别健康管理***，其特征在于，所述用户交互模块包括：用户个人信息管理、个人行为记录、建议提醒构成，为用户提供基本的交互操作，包括用户个人信息管理，展示其个人的行为的历史情况，展示建议提示等。
  3. 如权利要求1所述的基于传感器的分离式部署的人体行为识别健康管理***，其特征在于，所述数据采集模块对用户的行为数据进行采集，使用客户端所在硬件的传感器收集相应的数据，包括3轴加速度传感器数据、惯性传感器数据，并将采集到的数据进行预处理从而降低网络传输的数据量。
  4. 如权利要求1所述的基于传感器的分离式部署的人体行为识别健康管理***，其特征在于，所述模型识别模块采用机器学习的方法，利用服务端的计算硬件，将用户上传的数据进行快速准确的识别，并将识别的结果反馈给用户交互模块作为反馈，以及将识别结果传递给建议模块。
  5. 如权利要求1所述的基于传感器的分离式部署的人体行为识别健康管理***，其特征在于，所述数据分析模块根据用户的历史行为记录以及用户的个人状况分析用户过去一段时间的运动情况以及给出相应的建议。
  6. 如权利要求1所述的基于传感器的分离式部署的人体行为识别健康管理***，其特征在于，所述建议模块根据用户设置的个人状态对用户提出相应的建议，如运动量是否达到标准，某些运动是否过多应该减少，静坐时间过长应起身进行运动等。
PCT/CN2020/101145 2019-07-09 2020-07-09 基于传感器的分离式部署的人体行为识别健康管理*** WO2021004510A1 (zh)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN201910614552.4A CN110443145A (zh) 2019-07-09 2019-07-09 基于传感器的分离式部署的人体行为识别健康管理***
CN201910614552.4 2019-07-09

Publications (1)

Publication Number Publication Date
WO2021004510A1 true WO2021004510A1 (zh) 2021-01-14

Family

ID=68429946

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2020/101145 WO2021004510A1 (zh) 2019-07-09 2020-07-09 基于传感器的分离式部署的人体行为识别健康管理***

Country Status (2)

Country Link
CN (1) CN110443145A (zh)
WO (1) WO2021004510A1 (zh)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN116740813A (zh) * 2023-06-20 2023-09-12 深圳市视壮科技有限公司 一种基于ai图像识别行为监测的分析***及其方法

Families Citing this family (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN110443145A (zh) * 2019-07-09 2019-11-12 中山大学 基于传感器的分离式部署的人体行为识别健康管理***
CN111063437B (zh) * 2019-12-12 2024-01-23 中科海微(北京)科技有限公司 一种个性化慢病分析***
CN111700624B (zh) * 2020-07-27 2024-03-12 中国科学院合肥物质科学研究院 一种智能手环检测运动姿态的模式识别方法及***
CN112217837B (zh) * 2020-10-27 2023-07-14 常州信息职业技术学院 一种人体行为动作信息采集***

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20090075781A1 (en) * 2007-09-18 2009-03-19 Sensei, Inc. System for incorporating data from biometric devices into a feedback message to a mobile device
CN105095214A (zh) * 2014-04-22 2015-11-25 北京三星通信技术研究有限公司 基于运动识别进行信息推荐的方法及装置
CN105590022A (zh) * 2014-11-11 2016-05-18 宏达国际电子股份有限公司 身体状况建议方法及电子装置
CN109584989A (zh) * 2018-11-27 2019-04-05 北京羽扇智信息科技有限公司 一种运动提示信息的推送方法、装置、设备及存储介质
CN110443145A (zh) * 2019-07-09 2019-11-12 中山大学 基于传感器的分离式部署的人体行为识别健康管理***

Family Cites Families (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JP5322834B2 (ja) * 2009-08-06 2013-10-23 日本電信電話株式会社 情報表示システム及び情報表示方法

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20090075781A1 (en) * 2007-09-18 2009-03-19 Sensei, Inc. System for incorporating data from biometric devices into a feedback message to a mobile device
CN105095214A (zh) * 2014-04-22 2015-11-25 北京三星通信技术研究有限公司 基于运动识别进行信息推荐的方法及装置
CN105590022A (zh) * 2014-11-11 2016-05-18 宏达国际电子股份有限公司 身体状况建议方法及电子装置
CN109584989A (zh) * 2018-11-27 2019-04-05 北京羽扇智信息科技有限公司 一种运动提示信息的推送方法、装置、设备及存储介质
CN110443145A (zh) * 2019-07-09 2019-11-12 中山大学 基于传感器的分离式部署的人体行为识别健康管理***

Non-Patent Citations (2)

* Cited by examiner, † Cited by third party
Title
CHARISSA ANN RONAO ET AL.: "Human activity recognition with smartphone sensors using deep learning neural networks", EXPERT SYSTEMS WITH APPLICATIONS, 26 April 2016 (2016-04-26), XP029539400 *
SHENG, MING ET AL.: "A Moblie Health Service Platform Based on Big Data Analysis", SMART HEALTHCARE, no. 2,, 29 February 2016 (2016-02-29), ISSN: 2096-1219 *

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN116740813A (zh) * 2023-06-20 2023-09-12 深圳市视壮科技有限公司 一种基于ai图像识别行为监测的分析***及其方法
CN116740813B (zh) * 2023-06-20 2024-01-05 深圳市视壮科技有限公司 一种基于ai图像识别行为监测的分析***及其方法

Also Published As

Publication number Publication date
CN110443145A (zh) 2019-11-12

Similar Documents

Publication Publication Date Title
WO2021004510A1 (zh) 基于传感器的分离式部署的人体行为识别健康管理***
US20230333635A1 (en) Systems, methods, apparatuses and devices for detecting facial expression and for tracking movement and location in at least one of a virtual and augmented reality system
Xu et al. Emotion recognition from gait analyses: Current research and future directions
Vinola et al. A survey on human emotion recognition approaches, databases and applications
CN108062170A (zh) 基于卷积神经网络和智能终端的多类别人体姿态识别方法
KR102089002B1 (ko) 행동에 대한 피드백을 제공하는 웨어러블 디바이스 및 방법
Zhang et al. Analyzing students' attention in class using wearable devices
CN111414839A (zh) 基于姿态的情感识别方法及装置
Shin et al. Korean sign language recognition using EMG and IMU sensors based on group-dependent NN models
CN105787442A (zh) 一种基于视觉交互面向视障人群的穿戴式辅助***及其使用方法
CN112183314B (zh) 一种表情信息采集装置和表情识别方法及***
US11127181B2 (en) Avatar facial expression generating system and method of avatar facial expression generation
CN109976525A (zh) 一种用户界面交互方法、装置及计算机设备
WO2023216680A9 (zh) 一种数字生物标志物的形成***、形成方法及基于数字生物标志物的脑健康评判***
Nguyen et al. IMU-based spectrogram approach with deep convolutional neural networks for gait classification
CN109685148A (zh) 多类别人体动作识别方法及识别***
Kumar et al. Human activity recognition (har) using deep learning: Review, methodologies, progress and future research directions
CN110807471B (zh) 一种多模态传感器的行为识别***及识别方法
CN117064388A (zh) 基于情绪识别实现精神障碍评估分析的***
Bi et al. Smartge: identifying pen-holding gesture with smartwatch
Mallol-Ragolta et al. Outer product-based fusion of smartwatch sensor data for human activity recognition
CN112370058A (zh) 一种基于移动端对用户的情感进行识别与监测方法
CN113313795A (zh) 虚拟化身脸部表情生成***和虚拟化身脸部表情生成方法
CN116019443B (zh) 一种心肺复苏胸外按压合规性检测***及方法
CN107688828A (zh) 一种基于手机传感器的公交车拥挤程度估测方法

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 20837333

Country of ref document: EP

Kind code of ref document: A1

122 Ep: pct application non-entry in european phase

Ref document number: 20837333

Country of ref document: EP

Kind code of ref document: A1

122 Ep: pct application non-entry in european phase

Ref document number: 20837333

Country of ref document: EP

Kind code of ref document: A1

32PN Ep: public notification in the ep bulletin as address of the adressee cannot be established

Free format text: NOTING OF LOSS OF RIGHTS PURSUANT TO RULE 112(1) EPC (EPO FORM 1205A DATED 05.09.2022)

122 Ep: pct application non-entry in european phase

Ref document number: 20837333

Country of ref document: EP

Kind code of ref document: A1