Research on Coding and Decoding Technology of Tongue-computer Interface Based on EMG Signal Around the Ear

‍ ‍Human-computer interface （HCI） is a technology that converts human intentions or movements into instructions that the machine can understand. Among them， the HCI system， which uses bio-electric signals to realize real-time communication between human and external devices， can reflect the current internal state and expected action of a human， and has been widely used in many fields such as health monitoring， medical diagnosis， aerospace， prosthetics and auxiliary equipment development as well as many other fields. Some studies have shown that HCI system based on electromyography （EMG） signals has strong stability and high practicability， so it has broad application scenarios. Among these control modes， tongue movement is highly flexible and controllable， and the evoked signal has strong characteristics and can be detected easily， so tongue-computer interface （TCI）， which controls external devices through tongue movement， has extremely important research value. However， the existing research methods of tongue movement signal acquisition still cannot meet the requirements of high user comfort， precision correct classification accuracy and multiple control command set in natural scenes at the same time. To this end， seven different tongue movements were designed in this study， which were “left to right”， “right to left”， “up”， “down”， “stick”， “roll” and “speak ‘talk’”. The EMG signals of tongue movements in the non-hair area around the ear of 22 subjects were obtained by a more convenient and comfortable electrode placement method compared with previous research. In this research， the multi-dimensional information of tongue movement signals was extracted by time and frequency domain features and space domain features which was extracted by Common Spatial Pattern （CSP） algorithm， and seven tongue movement patterns were effectively classified by Support Vector Machines （SVM） algorithm. Finally， our research showed that the average classification accuracy of the seven types of tongue movement patterns of 22 subjects was up to 94.25%±5.23%. At the end of this research， we verifies the stability and separability of the EMG signals around the ear of tongue movements， which lays the foundation for the subsequent development of a high performance TCI system and the expansion of HCI application scenarios for the foreseeable future.