This paper proposes a new statistical approach, namely the probabilistic union model, for speech recognition subjected to unknown, time-varying, burst noise during the utterance. The model characterizes the partially and randomly corrupted observations based on the union of random events. We have tested the new model using the TIDIGITS database, corrupted by various type of additive abrupt noise. The experimental results show that the new model offers robustness to partial and temporal corruption, requiring little or no knowledge about the noise characteristics.
展开▼
