This research investigated whether acoustic-phonetic knowledge of the mother tongue of a non-native speaker can be used to adapt an existing target language phoneme HMM recognizer. For this purpose three sets of phoneme HMMs were generated, one representing the target language (German), one the mother tongue of the non-native speaker (Turkish), and the third the foreign-accented pronunciation of the target language (German spoken by Turkish speakers). The latter served as a benchmark for the tested adaptation methods. A derived Hidden Markov Model (HMM) clustering algorithm was applied on the target language phoneme HMM set using the mother tongue phoneme HMM set of the non-native speaker. Following the HMM adaptation a phoneme-level pronunciation technique was applied to generate phoneme mapping rules for the lexicon adaptation task. The results revealed a relative reduction of about 6% in WER for the adapted HMM. No further improvements were observed from the lexicon adaptation task.

In order to make hidden Markov model (HMM) speech recognition suitable for mobile phone applications, Siemens developed a recognizer, Very Smart Recognizer (VSR), for deployment in future mobile phone generations. Typical applications will be name dialling, command and control operations suited for different environments, for example in cars. The paper describes research and development issues of a speech recognizer in mobile devices focusing on noise robustness, memory efficiency and integer implementation. The VSR is shown to reach a word error rate as low as 4.1% on continuous digits recorded in a car environment. Furthermore by means of discriminative training and HMM-parameter coding, the memory requirements of the VSR HMMs are smaller than 64 kBytes.