Characteristic of the speaker independent speech recognition systems is the attribute that the user can immediately start voice recognition without a previous training phase. The vocabulary is limited to a few thousand words.
Speaker dependent speech recognition must be trained by the user prior to use for their own specific pronunciation. Use in applications with frequently changing users is not possible. The vocabulary is much larger than that of speaker independent recognizer in comparison.
In front-end system (also called on-line dictation), the processing of the reaction is carried out in text directly on the user, so that he can read the result substantially without significant time delay. In back-end systems (offline dictation or server-based detection), however, the reaction is performed on a remote server, thus, the text is rendered only with delay.
Such systems are particularly more common in the medical field. New distribution can be found on smartphones and similar mobile devices. One example is the application offered by an Online English Tutor.
Speaker independent speech recognition is preferably used in technical applications, for example in automatic dialogue systems such as a timetable. Elsewhere, where only a limited vocabulary is used, the speaker independent speech recognition is applied with success. It achieves detection of spoken English numbers from 0 to 9, at nearly 100% detection rate.
Very high recognition rates can be achieved in the use of speaker dependent speech recognition systems even on a limited vocabulary. However, where unlimited vocabulary is used, it does not ensure complete accuracy. Even an accuracy of 95 percent is too low.
In the meantime, current systems reach the dictates of personal computers running text on detection rates of about 99 percent, thus fulfilling the requirements for many areas of practice, such as for scientific texts, business correspondence, Online English Tutor or legal briefs.
Besides the size of the dictionary and the quality of the acoustic absorption plays a crucial role. For microphones that are mounted directly in front of the mouth (for example, headsets or phones) have a significantly higher recognition accuracy achieved than in more distant room mics.
However, important influencing factors are accurate pronunciation and related dictates of sufficiently long utterances, so that the language model can work optimally. To further increase the detection accuracy, the use of a video camera to shoot the face of the English tutor is often employed. The speech recognition systems then reads from the lip movements. By combining these results with the results of the acoustic detection, one can achieve a significantly higher detection rate especially in noisy environments.
Leave a Reply