Efficient low-order auto regressive moving average (ARMA) models for speech signals

Using model reduction, an efficient low order (ARMA) modeling process for speech is presented. In this approach, the modeling process starts with a relatively high order (AR) model obtained by some classical methods. The AR model is then reduced using the SVD-based method. The model reduction yields a reduced order ARMA model which interestingly preserves the key properties of the original full order model such as stability. Line spectral frequencies LSF and signal-to-noise ratio (SNR) behavior are also investigated. To illustrate the performance and the effectiveness of the proposed approach, some simulations are conducted on some practical speech segments, such as phonemes and sentences.