Abstract :
Reliable and efficient speech recognition has been one of the most elusive technology goals. For decades, people have been waiting for the day when they could toss away keyboards and other interface devices and communicate with computers, phones, and other systems in a way that´s as easy and natural as talking to another human being. Speech recognition has a long history, dating back to projects at Bell Labs long before the dawn of the computer age. Yet technology has so far largely failed to deliver the goods. For decades, developers´ attempts at producing usable speech recognition interfaces have failed to meet the mark in terms of usability and accuracy. But scientists and engineers at institutions worldwide are now taking advantage of sophisticated new technologies and fresh approaches to improve accuracy rates, boost conversion speeds, and make human-machine interactions as close to conversational as possible. The key to making speech recognition a mainstream technology is gaining a better understanding of the way people talk, listen, and comprehend sounds. Such investigations, many researchers believe, will lead to the development of more sophisticated algorithms to convert sounds into numbers as well as technologies designed to make speech recognition systems faster, more portable, and power efficient.
