A SPEECH –MAIL ASSISTANCE FOR VISIONLESS

Abstract

The technology innovations are successful with being accessible to the entire society. On observation of the problems faced by visionless, this paper presents speech supportive e-mailing system, such that the emails can be sent/read without any hassles. Although there are several commercial products for mailing purpose, they undergo accent ambiguity and thus lesser recognition accuracy rates. Besides this visionless people cannot verify the data and also the concept of homophones further reduces the accuracy rate of the system. On consideration of the aforesaid points, this paper conquers all the barriers by the inclusion of a customized dataset. The customer forms a customized dataset and it prompts the customer to utter alphabets one-by-one. Accuracy is the main factor for visionless, rather than time consumption. The proposed work extracts features from the customized dataset by means of Combined Feature Extraction Algorithm (CFEA) algorithm, which is a combination of MEL Frequency Cepstral Coefficient (MFCC), Linear Prediction Cepstral Coefficients (LPCC) and  RelAtive Spectral TrAnsform – Perceptual Linear Prediction (RASTA-PLP). Finally, k-NN classifier is employed to distinguish between the alphabets. The experimental results of the proposed work are satisfactory in terms of recognition accuracy rate.