Isolated Word Recognition Using Very Simple Recurrent Neural Network Plus and Noise Compensation LPC Analysis

In this paper, the isolated word recognition using very simple recurrent neural network plus (VSRN+) and noise compensation linear predictive coding (LPC) analysis is proposed. In the proposed system, a feature extraction based on noise compensation LPC analysis achieves accurate speech feature vector in noisy environment. Since VSRN+ is utilized as a recognizer, the proposed system is able to deal with complex time series data. Therefore, the proposed system yields accurate word recognition in noisy environment. From the investigation of performance evaluation, the proposed system improves word recognition rate in noisy environment.