Sensing and adaption for long-term hand rehabilitation

  • Peter Ralph Boyd

Student thesis: Doctoral Thesis


Following the loss of a limb, the success of the rehabilitative process for prosthesis control is dictated by the robustness of the prosthesis control scheme, the capabilities of the prosthesis, and the ease of the rehabilitative process. Unfortunately, limitations within current prosthesis sensing for hand rehabilitation are lacking and subsequently prosthesis rejection is still considerably high. Typical modalities in bio signal sensing focus on the usage of surface electromyography (sEMG) due to its nature of being non-invasive, intuitive, and effective. While the performance in laboratory settings have shown great promise, these methods typically are focused for intra-session or intra-day based efforts but are seldom applied to inter-day applications, leading to a reduction in the quality of life of amputees. To address lack of long term rehabilitation outcomes through bio signal sensing it is important to gain a deeper understanding into the nature of the transience within sEMG data and to expand into whether bio-signal led hand rehabilitation can be improved through fused sensing and adaption modalities. The contributions from this work therefore focus on how understanding and exploitation of intra-day sEMG signals can be exploited for long term use, the use of a-mode ultrasound sensor fusion for long term rehabilitation, and finally enacting closed loop haptic based hand control in a Virtual Environment based rehabilitation system.

Firstly, sEMG sensing strategies are proposed to support long term use with minimal patient input. Through the exploitation of transient changes within sEMG data during intra-day periods, a generalized system of long term hand motion recognition can be achieved. This thesis provides a deeper insight into the transient change of sEMG during frequent intra-day sensing by demonstrating the decay of hand motion recognition during a period of daily use. These transient changes are then exploited in training strategies that provide increase hand motion recognition accuracy for long term use.

Secondly, this thesis investigates the effect of sensor fusion led modalities towards long term hand motion recognition by means of a-mode ultrasound and sEMG. Unlike sEMG, a-mode ultrasound is capable of providing insight to deep muscle activity within the forearm while also being robust to crosstalk from neighboring muscle activation. The application of traditional sEMG features to a-mode ultrasound is evaluated when concerning sensor shift. Then a novel a-mode ultrasound led hand rehabilitation strategy is proposed through combination with sEMG sensing during large arm movement centred exercises demonstrating superior performance to sEMG alone.

Thirdly, in order to improve the feedback provided during rehabilitation and to enable closed loop hand control a virtual environment enhanced haptic rehabilitation system is proposed. In the proposed system, participants were provided feedback to their grasping effort through either visual force bars or electrotactile feedback while controlling a virtual hand. The performance of this rehabilitation strategy demonstrated a significantly reduced training period and highly repeatable finite hand control in various scenarios.
Date of Award2020
Original languageEnglish
Awarding Institution
  • University of Portsmouth
SupervisorZhaojie Ju (Supervisor), Nicholas John Savage (Supervisor) & Honghai Liu (Supervisor)

Cite this