Noninvasive sEMG-based Control for Humanoid Robot Teleoperated Navigation

Neural signal-based human-computer interfaces (HCIs) have been of huge interest and some cases have demonstrated the capability of the approaches successfully. Extracting users’ intentions through noninvasive sensory equipment provides information related to body movements faster than kinematic and dynamic sensory devices such as force sensors and motion trackers. Electromyography (EMG) is one of the most interesting neural signals to detect user’s intentions because it can be measured online achieving an acceptable signal-to noise ratio under current technologies. Surface EMG (sEMG) is measured on the skin noninvasively by electrical voltages which represent neuromuscular activities generated in muscles during their contraction. The activities are known to be discriminated according to intended behaviors. Some investigations have challenged to develop noninvasive sEMG-based interface techniques based on the fundamental properties of EMG. Some used EMG signals from eyebrows and jaw movements, from arms, and from legs. To cope with users’ desired commands (i.e., the direction of movements) in natural manner, sEMG signals from arms may be most appropriate because the interface protocol is intuitive enough to be familiarly operated. Furthermore, even the disabled and the elderly can use the sEMG-based interface to assess a controlled device without much training. Among the disabled, people with bilateral hand amputations and Spinal Cord Injuries (SCIs) at C6C7 functional levels could use their residual and controllable limbs. Due to its potential advantages, sEMG-based interfaces have been attempted to various applications: teleoperation, exoskeleton, computer interface, manipulation, fatigue estimation. However, there has been no attempt which applies a sEMG-based interface to control an independent agent such as a humanoid robot in a concrete practical scenario to the best of our knowledge. This work presents humanoid robot navigation control by extracting users’ intentions through light and wearable sEMG sensors and wireless remote communication between a user and a robot. Furthermore, we propose to integrate a finite state automation scheme to extend the number of discriminated intentions. To be practically usable, all processing procedures can be run in real-time without perceptible delay time. This paper is organized as follows. The components of the proposed method are explained in next section. Section 3 reports the experiment procedure and its results, and Section 4 concludes this paper. Noninvasive sEMG-based Control for Humanoid Robot Teleoperated Navigation