Evaluation of the Myo armband for the classification of hand motions

Pattern recognition-based control systems have been widely investigated in prostheses and virtual reality environments to improve amputees' quality of life. Most of these systems use surface electromyography (EMG) to detect user movement intentions. The Myo armband (MYB) is a wireless wearable device, developed by Thalmic Labs, which enables EMG recordings with a limited bandwidth (<100Hz). The aim of this study was to compare MYB's narrow bandwidth with a conventional EMG acquisition system (CONV) that captures the full EMG spectrum to assess its suitability for pattern recognition control. A crossover study was carried out with eight able-bodied participants, performing nine hand gestures. Six features were extracted from the data and classified by Linear Discriminant Analysis (LDA). Results showed a mean classification error of 5.82 ± 3.63% for CONV and 9.86 ± 8.05% for MYB with no significantly difference (P = 0.056). This implies that MYB may be suitable for pattern recognition applications despite the limitation in the bandwidth.