International Journal of Engineering and Advanced Technology (IJEAT) ISSN: 2249 – 8958, Volume-9 Issue-1S5, December, 2019 168 Retrieval Number: A10391291S52019/2019©BEIESP DOI: 10.35940/ijeat.A1039.1291S52019 Published By: Blue Eyes Intelligence Engineering & Sciences Publication Assessment of Wavelets Transform based Processing of Features of Forearm Muscle Signals for Prosthesis M. Karuna, Sitaramanjaneya Reddy Guntur Abstract- People who have lost forearm are suffer from hand mobility limitations due to trauma, disease or defect, Prosthesis arm help those people to do their daily actions. Researchers have been focused on developing artificial hand. In this regard, better processing of features of electromyographic (EMG) signal has a significant role from residual forearm muscle. To achieve this, Wavelet Transform (WT) technique has been applied because it is acceptable with the characteristics of EMG as a nonstationary signal. Results have shown that db5 wavelet decomposition performs best denoising at fifth level in other wavelets comparison. Furthermore, the ratio of Signal to Noise (S/N) and the error of percentage (PE) are calculated to evaluate the eminence and the usefulness features of EMG. Key Words: EMG, WT, Decomposition, Denoising, feature extraction, feature selection. I. INTRODUCTION EMG signal detected at the surface of the skin which determines the electrical current produced in fibres of muscle [1]. The main application of this research is to identify the various patterns of sEMG signal for controlling the prosthesis [2-3]. Noises are created in the EMG signal due to various sources such as the hardware for amplification, digital processing for analog to digital conversion and cables used for acquiring data as well as activity of motors at distance from detection area. Preprocessing of signal from muscle fibres acts an important role in realm of clinical and rehabilitation applications. Some methods to remove noise from the detected EMG signal have been emphasized by Cram et al. [4]. The major drawback of identifying the intentional movement is the inadequate consequences under the environment of presented noises, particularly when the random noise frequency characteristics. According to the literary sources, many researchers have suggested noise removal techniques from EMG signals by using digital filters [5]. Even though above filters can decrease the considerable noise, and also traces distortion in the EMG signal [6]. In recent research, the denoising WT theory is found very efficient in processing of denoise [7-10]. Therefore, signal decomposition, noise reduction from sEMG signal [11] using wavelets presented as shown in Fig.1. Moreover, an important requirement is to differentiate various EMG signals accurately for controlling prosthesis is effective extraction of features. Revised Manuscript Received on December 15, 2019. M. Karuna, Dept. of Electronics and Communication Engineering, Vignan Foundation for Science, Technology and Research, India. Sitaramanjaneya Reddy Guntur, Dept. of Electronics and Communication Engineering, Vignan Foundation for Science, Technology and Research, India. drgsr_ece@ vignan.ac.in The techniques based on the extraction of feature have been effectively used for recognising different forearm muscle movements [12]. Fig.1 Block diagram of wavelet denoising process In the present work were examined the effectiveness of denoising forearm muscle signals, with Stationary wavelet transform (SWT) and db5 at fifth level of decomposition of EMG signal by calculating the S/N values of the noise eliminated signals and Percentage Error (PE). In addition to that, a relative study was realized to picturize the efficiency of EMG features. The robustness of this approach depends on the better feature extraction. II. METHODOLOGY The four healthy male subjects were instructed to perform the wrist actions such as extension, flexion, pronation and supination. EMG detector used to collect EMG signals of forearm muscles, in which outputs for the signals, gain was adjusted to 60dB and bandwidth is limited to 20 Hz-500Hz with the help of main amplifier and filter.The sEMG signal was recorded by placing surface electrodes (Ag-AgCl) on the right forearm muscles such as flexor carpi radialis and extensor carpi radialis longus of a subject [13]. The equal distance of 2 cm is maintained between electrodes. One electrode is placed on the center of the muscle structure and other one is at the end. The third electrode was positioned on parts having no muscles on being bony. For each motion four datasets were collected. Recognitions of intentional movement through EMG signals have traced out markable results, since the last half a century period as a solution for dexterous prosthetic control to perform multifunctions.