Supervised neuro-fuzzy biofeedback for computer users

Abstract

The design of an advanced biofeedback system was introduced using neuro-fuzzy concept. Eleven healthy volunteers took part in six sessions over two weeks in which computer work was performed for 10 min. The six sessions were divided into two identical parts where each part was taken place in two consecutive days. After the first session, the subjects underwent excessive eccentric exercises of shoulder elevation to induce muscle soreness. The second session was performed immediately after the exercises and third session 24 hours after the exercises. The second part was performed exactly one week after the first part. Surface electromyography (EMG) of descending and ascending trapezius, deltoideus anterior and serratus anterior was recorded. Linear and nonlinear indices of muscular load were calculated from EMG signals. The first session was utilized as the benchmark of normal muscle condition during computer work and all the rest as mal-functioning/altered condition. A neuro-fuzzy system was trained and tested to discriminate between the first session and all the rest. Using a greedy forward search strategy most discriminative features were found. A high sensitivity ~90% but a low specificity ∼60% was observed. It was concluded that apart from the trapezius, the deltoideus and serratus anterior should also benefit from a biofeedback design. Combining of such system with timing constraint on biofeedback alarming can render a viable biofeedback system aiming at preventing musculoskeletal disorders. © 2011 Springer-Verlag.