Loquendo today released Loquendo ASR 7.7, the latest version of its speech recognition software that offers many new features to deliver robust and accurate speech recognition in 21 languages.
Among the enhancements are:
- Feature Normalization – an algorithm which improves recognition accuracy, even when run-time operating conditions are different from those on which the ASR was trained. It entails a run-time analysis and normalization of the audio signal.
- Continuous Recognition Mode – an automatic mechanism that enables the recognition of short, consecutive commands. Each command is recognised as separate, yet the ASR is immediately ready for the next recognition event.
- Improved Confidence Score algorithm – a robust algorithm for calculating confidence values. It improves separation between in-grammar and out-of-grammar utterances, resulting in more robust dialogues.
- Java support – built on C/C++ application programming interfaces.
Furthermore, the Acoustic Model Adaptation tool has been extended to support 16 kHz audio signals.
Loquendo ASR is available in 21 languages: U.S. and U.K. English, Russian, Argentinian, Chilean, Mexican and Castilian Spanish, Catalan, Valencian, Galician, French, Canadian French, Portuguese, Brazilian, Italian, Dutch, Polish, Turkish, German, Greek, and Swedish as well as Mixed Language Castilian Spanish and Catalan. Like all Loquendo products, Loquendo ASR is available for both Windows and Linux operating systems, and in the embedded version for all major embedded operating systems.