Fidelity estimation of speech transfered by delta modulator using neural network

This paper proposes a new method for speech signal fidelity estimation. The main tool is an Artifical Neural Network (ANN) having a capability to learn the words, recognize learned words, classify and compare them. Neural network inputs are coeflicient vectors derived fiom input signals. Paper presents a II class delta modulator computer software tool MATJAB 5.3. Simulations show changes in speech signal passing through delta modulator. The simulator is applied for automatic speech fidelity estimation using artifical neural nebvork without a human as an estimator.