Piano Teaching Action Recognition Based on LSTM

• Published in: Advances in Transdisciplinary Engineering
• Main Author: Sha R.; Tan B.
• Format: Conference paper
• Language: English
• Published: IOS Press BV 2024
• Online Access: https://www.scopus.com/inward/record.uri?eid=2-s2.0-85189515659&doi=10.3233%2fATDE231243&partnerID=40&md5=3a6c4c250a13dd8a52255b6981b75543

In this paper, a gesture recognition method based on Leap motion and LSTM is proposed to fully describe the establishment of finger touch motion adaptive curve during the process of piano playing teaching. The Leap motion sensor is used as the hardware platform to collect the finger touch action parameter values and establish gesture feature data. Then the dynamic gesture recognition is performed by integrating the Long Short-Term Memory networks, and the extracted single gesture action is further divided into frames. The optimal feature set and parameters are selected through experiments for further classification and recognition. The experimental results show that the multi feature recognition method based on LSTM can improve the recognition rate of similar gestures and the restoration accuracy is high, which can meet the requirements of action practice in piano playing teaching. © 2024 The Authors.