Automatic recognition of the language of spontaneous speech is a difficult problem that has been studied for years. This article provides a brief introduction to the general topic and an examination of a specific implementation that has already achieved a level of performance adequate for real applications. A detailed performance analysis provides a comparison of various techniques in terms of recognition accuracy and computational costs. Error rates of 5% and systems running eight-times real time have already been achieved. Issues of training on learning data and evaluation databases are addressed.
Click to purchase paper as a non-member or login as an AES member. If your company or school subscribes to the E-Library then switch to the institutional version. If you are not an AES member and would like to subscribe to the E-Library then Join the AES!
This paper costs $33 for non-members and is free for AES members and E-Library subscribers.