Word Detection and Global SNR Estimation of Speech Signals

Sreelakshmi M.S., Ann Nita Netto

Abstract: Speech recognition systems consists of machines or devices which are capable of receving speech as input and detecting the words or phrases of the input speech, converting the speech in to machine readable format (speech to text format) etc. But the introduction of noise to these systems may affect the performance of the system. The estimation of signal-to-noise ratio (SNR) gives an idea about the amount of noise present in the original signal. SNR compares the level of the data signal to the level of background noise. A novel method based on certain features of the speech signal is used for estimating the SNR. Combining the word detection scheme and estimation of global SNR improves the performance of such systems.

Keywords: Speech processing, MFCC, Signal-to-noise ratio, DNN, Word detection

How to Cite?: Sreelakshmi M.S., Ann Nita Netto, "Word Detection and Global SNR Estimation of Speech Signals", Volume 6 Issue 5, May 2017, International Journal of Science and Research (IJSR), Pages: 1362-1365, https://www.ijsr.net/getabstract.php?paperid=ART20173555, DOI: https://dx.doi.org/10.21275/ART20173555