Evaluation of muscle fatigue degree using surface electromyography and accelerometer signals in fall detection systems

Fall events are common to elderly people due to their deteriorating muscle structures caused by old age. Their relatively weaker bodies make them prone to accidents such as falls even when performing daily tasks. These fall events may leave physical or psychological consequences among them. Commonly, these events are associated with one or more identifiable risk factors such as weakness, unsteady gait, confusion, environment, and certain medications. Previous researches have shown that these events can be prevented using fall detection mechanisms. In this study, we investigate whether the analysis of muscle fatigue degree may enhance the performance of existing fall detection systems that utilize both surface electromyography (SEMG) and accelerometer (ACC) sensors. SEMG and ACC signals were measured and recorded from 20 healthy study volunteers. A series of pre-defined activities that mimic fall events were performed by the study volunteers. These activities were conducted in a controlled environment. Acquired SEMG signals were pre-processed to eliminate unwanted signals and distortion. Discriminative features were then extracted from the clean signals, and these extracted features were combined with the accelerometer data for classification using an Artificial Neural Network (ANN) classifier. Results showed that the combination of SEMG and ACC data have relatively increased the accuracy of fall detection systems.