Author(s)

Yin Chunyu

Corresponding Author

Yin Chunyu

Abstract

As an important technology in human-machine interaction, action recognition is one of the research hotspots in the pattern recognition. Accelerometer is a kind of wearable sensor. It is characterized by low energy consumption, small volume and low cost, which is widely used. At present, the action recognition process based on acceleration sensing is mainly divided into 3 steps: feature extraction, feature selection and recognition algorithm. In this paper, a recognition algorithm based on three-dimensional acceleration signal is proposed to identify different types of strokes according to the motion characteristics of people playing table tennis. The data used are collected from the accelerometer tied to the wrist and collected in the training process of table tennis. The threshold value of acceleration signal is judged by fixed threshold method. And the key characteristics of signal energy and gesture length are extracted.

Keywords

Acceleration sensor, Table tennis, Action recognition.