基于无声语音及肌电信息融合的多功能假肢控制研究

Most of the modern motorized prostheses are controlled with the electromyogram (EMG) from the residual muscles of the limbs. After amputation, the residual muscles are usually limited or completely lost, which can neither provide enough EMG for the multifunctional prostheses control, nor realize the flexible control of the prostheses. This project proposes to utilize the amputees' speech as additional control information and combine it together with myoelectric signal from the residual muscles for more flexible control of multifunctional prostheses. Considering the feeling of the users and the possible interference of the environment noise, the non-audible murmur (NAM) conducted by the soft tissue in the area of head and neck will be collected by a NAM transducer which is designed with condenser or piezoelectret to realize the accurate NAM recognition. Two strategies of NAM and EMG signal combination will be applied in the project: (1) By replacing the awful joint-switching method of the present EMG prostheses, the NAM recognition is used to select a joint of prosthetic arm. Then the EMG is used to decide a motion class involved in the selected joint; (2) The NAM recognition is used to select a motion class of multifunctional prostheses, and then the EMG is used as a startup signal of the prostheses. The successful implementation of this project will provide the important theoretical and technical supports for the improvement and enhancement of the control performance of a multifunctional prostheses system, and will also build the basics for the application of NAM sensing and recognition in other areas.

现代电动化假肢大多利用肢体残留肌肉产生的电信号进行控制，由于截肢后肢体残留肌肉有限或完全丧失，不能为实现多功能假肢控制提供足够的肌电信息，也难以实现假肢的灵活控制。本项目拟利用截肢者的语音作为假肢的控制信息，并与残留肌肉产生的肌电信息融合，研究如何实现多功能假肢的灵活控制。考虑到假肢使用者的感受及环境噪声干扰，将通过设计电容式及压电式无声语音传感器，采集使用者头颈部区域经软组织传导的无声带振动语音信号，实现无声语音信号的精确识别。采用语音与肌电信息的两种融合模式：（1）用无声语音识别替代现有肌电假肢笨拙的关节切换方式，实现关节选择，进而用肌电信号确定动作的类型；（2）用无声语音识别实现多功能假肢动作类型的选择，进而用肌电信号启动动作的执行。本项目的成功实施将为改善和提高多功能假肢系统的操控性能提供理论与技术支持，并为无声语音传感与识别在其它领域的应用提供研究基础。