EMG-Based Noncontact Human-Computer Interface for Letter and Character Inputting

摘要

Human-computer interface (HCI) is an important way for information transmission between human and computer. This study aims to design a kind of noncontact HCI. Handwriting recognition for computer input is realized by decoding surface electromyography (sEMG) signals from users. In terms of signal processing, a sample entropy-based segmentation algorithm, normalized processing and vector quantization, and hidden Markov model (HMM), are proposed. The from-left-to-right method is used to determine the initial HMM parameters. The continuous inputting of letters and characters is realized with the help of AEVIOUS virtual sliding keyboard, and the average online recognition accuracy on four subjects is 91.8% for 10 numbers and 87.6% for 26 letters.