Speech modelling by model-order reduction: SNR behaviour

Using model reduction, a new approach for low-order speech modelling is presented. The process starts with a relatively high-order (full-order) autoregressive (AR) model obtained by some classical methods. The AR model is then reduced using the state projection method. The model reduction yields a reduced-order autoregressive moving-average model which interestingly preserves the key properties of the original model such as stability and minimality. Phase minimality and SNR behaviour are investigated. To illustrate the performance and the effectiveness of the proposed approach, computer simulations are conducted on practical speech segments.