Physiological Interaction - Automatic Speech Recognition

Automatic Speech Recognition

The automatic speech recognition (ASR) system will capture spoken words and then classify word groupings to form a sentence. ASR is computer-driven and transcribes spoken language into text that can be read by using real time. Special applications for this system may be used to help deaf and hard of hearing people understand what is being said. (NETAC, 1998) This technology is becoming more popular in computing environments as well. Call centres are taking on automatic speech recognition in order to decrease the amount of clients who hang up because they are not talking to a live operator. (Haag, 2006)

In order for an ASR system to function, it must follow three steps. The system will capture words that are spoken into a microphone, while taking out any background noises, and then converting the digital signals of the speech into syllables (phonemes). This is referred to as feature analysis. Next, the system will match the spoken syllables to a phoneme sequence that is kept in an acoustic model database. This is called pattern classification. The system will then try to make sense of what is being said by comparing the word phonemes from the previous step, this time with a language model database. (Haag, 2006)