This paper describes the proposal of indigenous development of a technological platform for man-machine interaction with the prosthetic hand using a surface electromyogram (sEMG) signals in the laboratory. This work involves the design and the development of the sEMG conditioning system, implementation of EMG pattern recognition algorithm using TMS320F28335 digital signal controller (DSC) and generation of control signals for actuation of hand. In this work, EMG signals were acquired from ten healthy subjects and two transradial amputees for the six hand motions in off-line. The time domain features were extracted from each channel was grouped into five ensembles to understand the importance of selection of features in the identification of intention from EMG signals. Further, the performance of different classifiers namely simple logistic regression (SLR), J48 algorithm for decision tree (DT), logistic model tree (LMT), neural network (NN) and linear discriminant analysis (LDA) were studied. The Kruscal-Wallis test is performed for classification accuracy, computation time and memory space required for different feature ensembles to identify the effective feature ensemble. Also the performance of the classifier was tested in on-line with transradial amputees for actuation of prosthetic hand for two intended motions with TMS320F28335 controller using efficient ensemble and classifier. © 2014 Elsevier B.V. All rights reserved.