Revisiting the Utility of Spectral Measures on Spoken Letter Recognition

For more than a century, speech and audio processing have been essential components of entertainment and telecommunication systems. The content of recorded sound signals can be examined, understood, and interpreted by technicians using several audio analysis techniques. Some result in visualization...

Full description

Saved in:
Bibliographic Details
Main Authors: Asokan, J., Dheeraj, Gurazada Sai Siva Naga, Naidu, Ravilisetty Venkata Satya Sai Mastan, Reddy, Ranamalla Shriya, Sasi, G., Elamaran, V.
Format: Conference Proceeding
Language:English
Subjects:
Entertainment industry
FFT
Measurement techniques
Recording
spectral measures
Speech processing
Speech recognition
speech signal
Telecommunications
unvoiced consonants
Visualization
voiced consonants
Online Access:Request full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:For more than a century, speech and audio processing have been essential components of entertainment and telecommunication systems. The content of recorded sound signals can be examined, understood, and interpreted by technicians using several audio analysis techniques. Some result in visualization techniques, while others may help specify additional audio signal processing or measurement. The spoken letters in this study are identified using a straightforward spectrum comparison or spectral measurements technique. Additionally, by computing the FFT of the recording, the low and high-frequency sums, and finally, the ratio, we can reasonably predict whether an unknown recording of a single letter is a "C" or an "R". Finally, the spectral measures technique is performed to detect the unvoiced consonants (c, f, s, t) versus the voiced consonants (b, d, l, m, n, r, y) is performed using the spectral measures technique. The Matlab 9.4 tool is used for experimental simulations with speech signals.
ISSN:2575-7288
DOI:10.1109/ICACCS57279.2023.10112969