Automatic Speech Recognition System Dedicated for Polish

An automatic speech recognition system for Polish is demonstrated. A few layers of our system are different from popular approaches as a result of differences between Polish and English languages. Research on automatic speech recognition (ASR) started several decades ago. Most of the progress in the field was done for English. It has resulted in many successful designs, however ASR systems are always below the level of human speech recognition capability, even for English. In case of less popular languages, like Polish (with around 60 million speakers), the situation is much worse. There is no large vocabulary ASR (LVR) software for Polish. Polish speech contains very high-frequency phones (fricatives and plosives) and the language is highly inflected and non-positional. There are some commercial call centre applications, developed by PrimeSpeech, but they are limited to their domain areas. Our system is based on modified kNN classifier and wavelets. It is targeted for Polish, while others [1, 6, 7, 5] are more general, and strongly based on HTK framework [9].