A research of SMG controlled prosthetic hand with SSE2 acceleration

The muscle thickness changes of forearm extensor extracted from ultrasound images have been used to control prosthetic hand in our previous work, but the execution time to calculate the muscle thickness changes is a serious problem. In this study, the sum of absolute difference (SAD) algorithm was used instead of cross-correlation (CC) algorithm to track the muscle deformation. What's more, the Streaming SIMD Extensions 2 (SSE2) technique was used to further optimize the SAD algorithm. The results illuminated that the SAD with SSE2 drastically reduced mean 60% execution time comparing with CC algorithm, and the frame rate of 13.48 Hz was greater than the ultrasound sample rate of 12 Hz. So the prosthetic hand could be well controlled in a real-time way by this fast tracking algorithm. This study suggests that the muscle deformation extracted from ultrasound image has the feasibility to control prosthetic hand.