Digital Processing of Speech Signals

Abstract

This thesis represents an efficient method for reducing the quantization noise for coding speech signals at low bit rates. This method has gained interest due to its ease of implementation and channel-noise robustness compared to other digital Coders. Adaptive Differential Pulse Code Modulation (ADPCM) is one of the most efficient schemes for digital encoding of speech. Different algorithms have been proposed for increasing the dynamic range for a given value of Signal-to-quantization Noise Ratio (SNR). The predictor is a vital part of the overall system. This thesis also investigates the use of second­ order prediction to increase the SNR for several ADPCM systems at low bit rates spechialy Jayant Adaptive Quantizer (JAQ) and Icremental Adaptive Quantizer (IAQ). Inaddition a study and comparsion has been done between the different adaptation systems. Moreover computer simulations have been performed using three different types of input signals, sinusoidal signal, random signal and real speech signal. Inaddition, a unit step function is used to show that the IAQ is the more suitable to follow the variation for speech signal. It concludes that the second order prediction gives better performance than the first order predication. Four decibels improvement in SNR has been achieved in the case of real speech signal and consequently the reduction bit rate has been obtained. Furthermore the hardware implementation of this system is constructed to realize the predictive coding with good digital accuracy which reduced word size.