Application of model order reduction approach on quality evaluation of speech signal

In this paper, a model order reduction approach has been used to compare the performance of hearing quality of the voice sound of lower and higher order systems. To implement this, an input signal in the form of one of the standard speech signal from IPA database has been applied on the original higher order system as well as on the reduced order system. This form of the input signal is applied to generate the voice sound of the speech signal and a comparison is being made on the hearing quality of the output of both the systems. First of all, one of the model reduction approach i.e. Schur approach has been applied on the original higher order system to obtain a reduced order system by eliminating undesired order without losing any key properties of original higher order system. Now, the reduced order system contains the lower order as desired by the user. After that, an input speech signal is applied separately on both the systems. Then, the outputs of both the systems have been obtained by the convolution operation. Further, the output signal of original higher order system in distorted form and reconstructed signal of the lower order system are obtained by de-convolution operation separately. Finally, a comparison has been made on the output signals of both the systems and observed that the voice sound of generated speech signal by lower order system has good hearing quality. But the generated voice sound by original higher order system has lost its quality. So, a model order reduction approach gives far much better hearing quality of the voice sound by the processing through lower order system in comparison with the original higher order system.