Feature extraction method and system for human motion, recognition method and system

Abstract

The invention relates to a feature extraction method and system for human body movement. The method comprises the following steps: collecting signal data generated during human body movement; Extracting heterosexual information in the signal data, the heterosexual information including muscle external motion trajectory signal data information and muscle internal vibration signal data information;performing A feature extraction for the heterogeneous information, wherein the feature includes an external muscle motion trajectory signal feature, an internal muscle vibration signal feature, and acombination feature, and the combination feature refers to an associated feature obtained from the external muscle motion trajectory signal data information and the internal muscle vibration signal data information.

Description

technical field [0001] The present invention relates to the computing field, in particular to a feature extraction method and system for human motion, and a recognition method and system. Background technique [0002] Human motion recognition is a combination of sensor data collection and data mining technologies to realize the recognition and recording of human behavior. Human motion recognition has a wide range of applications, such as the recognition of routine life behaviors such as eating, watching TV, and rest, or gait recognition such as walking, running, and climbing stairs. Human motion recognition can bring many potential benefits and advantages to people's lives, for example, recording and monitoring exercise fitness, people's own health management or assisting in predicting behavioral intentions, etc., has been a research hotspot for scholars for many years. [0003] Generally speaking, there are many ways to classify human body movements that can be used as rec...

AI Technical Summary

Problems solved by technology

[0004] Human body movement recognition methods in the prior art can be divided into the following two types, one is the recognition method based on non-wearable devices, for example, using a camera to record the movement of the human body, and then using the method of image recognition to analyze the human body movement collected by the camera This method is limited by the acquisition conditions and may lead to inaccurate data acquisition, and compared with the method of processing sensor acquisition data, the data processing for image recognition is also more complicated; another identification method is based on The recognition method of wearable devices. This method can use a variety of wearable devices, such as accelerometer sensors with low cost. Although such wearable devices have high recognition accuracy for non-kinship actions, their recognition accuracy for kinship actions is relatively high. However, the sensors for measuring physiological signals (such as myoelectric electrodes and heart rate sensors) have poor overall recognition accuracy, and the cost of the myoelectric electrodes is relatively high, which is not conducive to popularization and use; The combination of sensors (such as the combination of accelerometer and heart rate sensor) will lead to too many sensors worn by customers and poor experience

Images