Robust speech recognition algorithms in a car noise environment

In this paper, we present several new robust isolated word speech recognition systems which employ FMQ/MQ as the spectral labelling process, followed by a Hidden Markov Model (HMM), or a HMM and Neural Network (HMM/MLP) classification technique. The ISWR systems provide selective input data to a neural network in response to speech signal to acoustic noise ratios to improve speech recognition system performance. Simply, FMQ/HMM system can exploit error compensation from FVQ/HMM processes. TIDIGITS and NOSEX_92 [7] have been used as the speech and noise databases. These robust algorithms ensure a high recognition accuracy performance even at input SNR as low as 5 and 0 dBs.