Abstract: This paper presents a new multi pose audio visual speech recognition system based on fusion of side pose visual features and acoustic signals. The proposed method improved robustness and circumvention of conventional multimodal speech recognition system. The work was implemented on ‘vVISWA’ (Visual Vocabulary of Independent Standard Words) dataset comprised of full frontal, 45degree and side pose visual streams.The feature sets originating from the visual feature for Side pose are extracted using 2D Stationary Wavelet Transform (2D-SWT) and acoustic features extracted using (Linear Predictive Coding) LPC were fused and classified using KNN algorithm resulted in 90 % accuracy. This work facilitates approach of automatic recognition of isolated words from side pose in Multipose audio visual speech recognition domain where partial visual features of face were exists. Keywords: Side pose face detection, stationary wavelet transform, linear predictive analysis, Feature level fusion, KNN classifier.