Configuration and analysis of piezoelectric-based in socket sensory system for transfemoral prosthetic Gait detection / Farahiyah Jasni

Farahiyah , Jasni (2018) Configuration and analysis of piezoelectric-based in socket sensory system for transfemoral prosthetic Gait detection / Farahiyah Jasni. PhD thesis, University of Malaya.

	PDF (The Candidate's Agreement) Restricted to Repository staff only Download (1643Kb)
Preview	PDF (Thesis PhD) Download (3423Kb) | Preview

Abstract

Sensory system is one of the important components in micro-processor controlled (MPC) prosthetic leg. It provides the information to the controller on the current phase of the device and, in some applications, the signals from the sensory system are used to predict the intended movement of the amputee. Hence, the quality of the sensory system for the MPC prosthetic leg is very important. The challenge is to balance between the practicality and the efficiency of the sensory system. This study investigated the feasibility and efficacy of an in-socket sensory system which uses multiple piezoelectric sensors as a sensory system for transfemoral (TF) MPC prosthetic leg. The methodology comprised of selecting: (a) the best piezoelectric sensor to be used in terms of the material, size and shape, (b) the method of mounting the sensors onto the socket, and (c) the placement of the sensors. The finalized configuration that resulted in the best response was adopted and used to collect data for training and classifying the gait phase using pattern recognition method. The performance of the proposed sensory system was evaluated by checking the accuracy of the pattern recognition algorithm in detecting the gait phases at different speeds of normal walking. Findings from the simulation study suggested that Polyvinylidene Fluoride (PVDF)-based, small rectangular piezoelectric sensor yielded the highest output voltage. The cantilever with elastic foundation mounting configuration should be adopted and the sensors should be positioned in a zig-zag orientation from top to bottom on the anterior and posterior socket wall to cover the quadriceps and hamstring muscle groups. The sensory system that has been configured was tested and consistent signals across multiple trials were found as the average Frobenius norm, ||A||F, calculated for all sensors was 4.5 out of 5.0. The response from the sensory system was proven to be feasible to be used to identify TF gait phases and it could be used to characterise different walking types or speed. For the analysis of gait detection using pattern recognition method through signals from the novel sensory system, it can be concluded that the ensemble classifier method, with a window size of 100 ms, produced the best training and testing performance with an average of 95.6% and 88.32% classification accuracy, respectively. The accuracy increased by 0.1% when the number of sensors was optimized to 14 sensors. In conclusion, this thesis presents the configuration for the in-socket sensory system for TF prosthesis, and proved that the proposed system is feasible to be used to detect gait phases for level walking. In addition, this thesis offers the foundation for more studies on utilization of in-socket sensory system for the MPC prosthetic leg control.

Actions (For repository staff only : Login required)