APLIKASI PENGENALAN UCAPAN DENGAN JARINGAN SYARAF TIRUAN PROPAGASI BALIK UNTUK PENGENDALIAN SMART WHEELCHAIR

Abstract

Speech recognition process can be done in many ways one of them with artificial neural networks. In order to be easily understood and to understand, it would require some method of characteristics extraction methods such as by LPC and Fourier transformation. Linear Predictive Coding is one tool in signal processing or signal analysis to obtain the unique characteristics of each sound pattern. While the Fourier transform is used to clarify the characteristics of each pattern as it can provide information that is presented in the frequency domain of both discrete and continuous. ANN (Artificial Neural Networks) are widely used for various applications of pattern recognition. The ability of learning from training data and generalize to the situation / condition is new is the fundamental reason why ANN is used. In this Final Project, ANN used is the Back Propagation. Furthermore, speech recognition results are translated into ASCII characters are then sent to a wheelchair through a serial connection that was identified as five voice commands. The voice commands consist of commands forward, backward, right, left, stop. Test results show that the recognition of new data lower then the exercise data. From some variation of test, obtained the best network tes_fix3, with speech recognition percentage of respondents in the database is 87%, and speech recognition percentage of respondents outside the database is 85,33%. Key Word : speech recognition, LPC, Fourier, ANN, Back Propagation, serial, wheelchair