Automatic Language Identification of Telephone Speech

II Lincoln Laboratory has investigated the development of a system that can automatically identify the language of a speech utterance. To perform the task of automatic language identification, we have experimented with four approaches: Gaussian mixture model classification; single-language phone recognition followed by language modeling (PRLM); parallel PRLM, which uses multiple single-language phone recognizers, each trained in a different language; and language-dependent parallel phone recognition. These four approaches, which span a wide range of training requirements and levels of recognition complexity, were evaluated with the Oregon Graduate Institute Multi-Language Telephone Speech Corpus. Our results show that the three systems with phone recognizers achieved higher performance than the simpler Gaussian mixture classifier. The top-performing system was parallel PRLM, which performed two-language, closed-set, forced-choice classification with a 2% error rate for 45-sec utterances and a 5% error rate for lO-sec utterances. For eleven-language classification, parallel PRLM exhibited an 11% error rate for 45-sec utterances and a 21% error rate for 10-sec utterances.