Abstract:
Speech compression is one area of digital signal processing that can be used to convert human speech into an efficient encoded representation that again can be decoded to produce a close approximation. Our main aim of this work is to compress the recorded speech. Speech coding is a lossy type of coding, which means that the output signal does not sound like the input. Now a days speech compression has many applications in the area of telecommunications such as digital cellular phones, voice mail and internet phones and also in high quality speech storage and message encryption.
In this dissertation work three speech compression techniques has been implemented for the purpose of compressing the recorded speech. These are Adaptive Differential Pulse Code Modulation, Linear Predictive Coding and Wavelet Transform. Out of these three, Wavelet Transform Based Speech Compression is a new technique. Wavelets have been successfully used in image compression applications. But less attention has been paid towards its application in the field of the speech compression. The aim of this dissertation work has been centered around implementation and comparison of speech compression techniques. Our comparative evaluation was based on the following parameters.
1. Compression Ratio
2. Signal to Noise Ratio
3. Peak Signal to Noise Ratio
4. Normalized Root Mean Square Error
In this work it is proved that compression ratio in the case of ADPCM and LPC is not variable where as in the case of wavelet transform based speech compression, compression ratio is variable and quality is also quite good with respect to ADPCM and LPC. It is also concluded that quality decreases by increasing the compression ratio. In the case of ADPCM compression ratio is less but the quality is good. In the case of LPC compression ratio is more but the quality is poor. Quality and compression ratio
