I love cv's http://www.ling.gu.se/~jimh/courses/ShortResume.html Research and implementation of multilanguage speech recognition and a spoken language identification system. Designed and implemented a structured approach to spoken language identification using Hidden Markov Model phoneme recognition with phonotactic constraints with colleague, Shubha Kadambe. Our best performing system used a complete large vocabulary speech recognition system for each language with the Bayes equation including the probability of acoustics in the denominator. The language with the best probability after 10 seconds of speech was the language chosen. The performance was 95% correct language ID with 5 languages (Mandarin Chinese, German, Spanish, Japanese and English). The system was tested as part of the NSA Language Identification Program. A patent was filed this technique. Cheers, Julian.
