Stumble detection and classification for an intelligent transfemoral prosthesis

This paper describes an approach for the real-time detection of stumble for use in an intelligent lower limb prosthesis, using accelerometers mounted on the prosthesis, and also describes an algorithm that classifies the stumble response as either an elevating or lowering type response. In order to validate the proposed approach, the investigators collected stumble data on 10 healthy subjects using accelerometers affixed to the subjects in a manner consistent with similar instrumentation on a transfemoral prosthesis. The proposed algorithms were shown to correctly identify stumbling and correctly classify the stumble response for all 19 stumbles and 34 control strides collected in the experiments.