Missing data mask models with global frequency and temporal constraints

Missing data recognition has been developed in order to increase noise robustness in automatic speech recognition. Many differ-ent factors, including the speech decoding process itself, shall be considered to locate the masks. In this work, we are considering Bayesian models of the masks, where every spectral feature is clas-sified as reliable or masked, and is independent from the rest of the signal. This classification strategy can produce unrelated small “spots”, while experiments suggest that oracle reliable and unreli-able features tend to be clustered into time-frequency blocks. We call this undesired effect: the “checkerboard ” effect. In this paper, we propose a new Bayesian missing data classi-fier that integrates frequency and temporal constraints in order to reduce, or avoid, this “checkerboard ” effect. The proposed clas-sifier is evaluated on the Aurora2 connected digit corpora. Inte-grating such constraints in the missing data classification leads to significant improvements in recognition accuracy. Index Terms: missing data recognition, mask estimation, tempo-ral constraints, frequency constraints