The creation of ASR system had an aim to provide people with a machine that can translate each word spoken by them in real time. The system still has enough to incorporate and advance more. There still is a gap between expectations from ASR systems and their performance. The ASR systems are meant to recognize the speech of human beings without being interrupted from the speaker’s accent, choice of words, noise or other features. Commercially available ASR systems need a lesser amount of speaker training and have the capability to recognize the continuous speech vocabulary with higher efficiency along with a broader range of vocabularies being captured. Commercial companies are often found to claim that ASR systems provide 98 to 99 percent accuracy provided that they are working under optimal conditions. Optimal conditions are when the users have speech habits which are in line with the training data, when the users have proper speaker adaptation and when the process is carried out in a noise-free environment. It can be explained from this information that recognition rates for heavily accented people may be lower than others.
Although there still is a room for improvement in Automatic Speech Recognition (ASR) systems, there are a number of application areas which benefit from its use. Telecommunications is one of the major application areas as speech recognition software acts as an interface that directly transfers data through a communication system into the information system. Today, inquiry systems, dialing assistance and telephony interpretations are some of the examples where ASR systems are used. Office automation is another area that benefits from the use of ASR. The crucial application areas of ASR include ASR in CAD applications and providing input using direct command in computers. Medical applications also make use of this technology for