Gesture identification method based on acceleration transducer

Abstract

The invention discloses a gesture identification method based on an acceleration transducer, which comprises the following steps: (1) detecting and recording a three-dimensional acceleration data sequence of a user's gesture by a handheld device; (2) framing the three-dimensional acceleration data sequence; (3) extracting intra-frame feature descriptors; (4) forming overall feature descriptors; (5) training on the basis of a gesture data set model of a support vector machine; and (6) identifying actions on the basis of the support vector machine. The invention reduces the dependence on the environment and the user in the process of gesture identification and also lightens the effects of noises and changes in the acceleration signals, thereby enhancing the performance of gesture identification.

Description

technical field [0001] The invention relates to a gesture recognition method based on an acceleration sensor, in particular to a method for analyzing three-dimensional acceleration data sequences by using frame division processing and extracting time domain and frequency domain signal features within the frame, and using multi-class support vector machines as models to perform Methods for Gesture Classification and Recognition. Background technique [0002] With the increasing application of electrical appliances and computing devices in daily life, promoting the development of new human-computer interaction methods in various modalities has become a hotspot in computing technology research. Among them, the way of interacting with the device by the user's dynamic gesture has broad application prospects, especially in the fields of household appliance control, handheld device (such as mobile phone) control and the like. [0003] Traditional gesture recognition is realized ba...

AI Technical Summary

Problems solved by technology

This technology has a strong dependence on the light of the surrounding environment and the position and direction of the user, which obviously cannot meet the actual requirements of the user's changeable daily use

The existing gesture recognition methods based on acceleration sensors mostly use the time-domain information of the acceleration signal during the gesture process, such as hidden Markov model, dynamic time warping model, etc., which make the recognition process very sensitive to signal noise and signal changes. The accuracy of gesture recognition

Images