Speech recognition under musical environments using kalman filter and iterative MLLR adaptation

In this paper, we propose a speech recognition method under non-stationary musical environments using Kalman ltering speech signal estimation method and iterative unsupervised MLLR(Maximum Likelihood Linear Regression) adaptation. Our proposing method estimates the speech signal under non-stationary noisy environments such a s m usical background by applying speech state transition model to Kalman ltering estimation. The speech state transition model represents the state transition of speech component in non-stationary noisy speech and is modeled by using Taylor expansion. In this model, the state transition of noise component is estimated by using linear predictive estimation. Furthermore , to obtain higher recognition accuracy, w e consider to adapt the acoustic models by using iterative unsuper-vised MLLR adaptation to speech spectra distorted by Kalman ltering residual noise. In order to evaluate the proposed method, we carried out large vocabulary continuous speech recognition experiments under 3 types of music. As a result, the proposed method obtained the signiicant improvement in word accuracy, from 20.04% to 64.43% at 0dB SNR.