Surface Electromyography based Pakistani Sign Language Interpreter

Electromyography signal is the electrical activity of the signals that quantitatively measured and it provides a significant amount of information related to various muscular activities. This research focuses on surface Electromyography (sEMG) technique for the detection of electrical signals that are produced as a result of hand movements for various alphabetical signs of Pakistani Sign Language (PSL). The data set was gathered using the BIOPAC system and 30 signals per alphabet were collected giving us a total of 26 classes. The pre-processing of data was established using Empirical Mode Decomposition and a total of 15 features were extracted for developing distinct characteristics in discrete classes. Different features from the time domain, spectral domain, statistical domain, and many other domains were fed to Linear Discriminant Classifier for the classification of data set. Statistical evaluation parameters were utilized to validate the results obtained through classification along 10-fold cross-validation. The results obtained from the study show an accuracy of 81.0%, a sensitivity of 84.08%, and specificity of 84.7%. Moreover, this study serves as a stepping stone towards the use of EMG for the interpretation of Pakistani Sign Language.