BUT OpenSAT 2017 Speech Recognition System

The paper describes BUT Automatic Speech Recognition (ASR) systems for two domains in OpenSAT evaluations: Low Resourced Languages and Public Safety Communications. The first was challenging due to lack of training data, therefore multilingual approaches for BLSTM training were employed and recently published Residual Memory Networks requiring less training data were used. Combination of both approaches led to superior performance. The second domain was challenging due to recording in extreme conditions: specific channel, speaker under stress, high levels of noise. A data augmentation process was very important to get reasonably good performance.