Web Services Based Hybrid Recognizer of Lithuanian Voice Commands

Abstract

This paper presents the recently developed medical-pharmaceutical informative system with voice user interface. This is the first computerized system oriented towards healthcare services and industry where Lithuanian voice commands are used as a primary mean for control. Another essential property of the developed system is its hybrid nature: two different recognizers -an adapted commercial Spanish speech recognizer available from Microsoft and a locally developed HMM speech recognizer based on Lithuanian acoustic models -are operating in parallel. The recognition hypotheses produced by those recognizers are joined together using logical rules obtained using decision rules induction algorithms such as Ripper. All these measures and approaches allowed achieve very high speaker independent voice commands recognition accuracy acceptable for the system implementation in practice. The best achieved recognition was 98.9 % for 1000 Lithuanian voice commands. The paper presents optimization issues related with the development of the system.

Some Aspects of Voice User Interfaces Development for Internet And Computer Control Applications

“…This study is the continuation of our earlier research. In [8] we showed that Microsoft SAPI could be applied to recognize Lithuanian voice commands. In [4] method to select transcriptions to recognize Lithuanian commands using foreign language speech engine was shown.…”

Evaluation of Methods to Combine Different Speech Recognizers

“…At the same time it has been shown that proper adaptation of existing foreign language acoustic models could speed up the development of recognizer and lead to the acceptable recognition level in that language [1]- [4], [6], [7]. Some previous studies have shown that speech recognition systems of languages such as English, Spanish or Russian can be quite well adapted for Lithuanian speech recognition [1], [3], [4]. However, the recognition results are not always as good as necessary and depend on many factors.…”

“…What is more, researchers are experimenting with different acoustic models adaptation methods in order to maximize the recognition performance with small amount of non-native data available [11]. Statistical algorithms for combining different acoustic models are used quite often and produces promising results [1], [3], [4], [6], [11], [12]. These researches shows, that in many cases it is possible to achieve high enough recognition accuracy by using hybrid systems with adapted acoustic models.…”