An efficient low-bit rate CELP speech coding algorithm

Voice communication is one of the easiest and natural methods of communication between human beings. In recent times it became popular because of mobile communications. It contains information as well as personal information such as age, identity, emotions etc. The requirement for efficient coding method is still a research topic for noise environment. Speech coding technique reduces the bandwidth requirements and storage space for speech data. Efficient coding method helps to reduce the bit rate and at the same time keep the speech quality and intelligibility at reasonable rate. Many speech coders have been developed to compress effectively and to ensure optimized performance to the human ear. Service providers of mobile phone communications are looking for low bit rate coders to allocate more number of users in limited bandwidth. In this paper a mixed coding method i.e. Code Exited Linear Prediction (CELP) is implemented on noise free speech signals using MATLAB. With this more exact modeling of spectral zeros is possible compared to the linear predictive coding (LPC). Subjective tests indicated that the coder at 16kbps and 9.6kbps achieves a significant improvement in performance over LPC coder under the same coding framework and bit allocation.