A 13 kb/s wideband CELP speech encoder was developped. This is an area of increasing growth and interest due to some emerging applications like :multimedia devices, videoconferencing, ISDN applications, etc; these scenarios require high-quality speech without the constraint of the limited telephonic bandwidth. Thus, the bandwidth considered in those applications goes from very low frequencies (around 50 Hz) up to 7000 Hz. The sampling frequency typically used is 16 kHz, although higher sampling frequencies are under consideration for some applications. The research goal consists of reducing the bit rate while maintaining the subjective quality. One way to approach the problem is to extend the telephonic bandwidth schemes to this scenario, tuning them to handle chiefly speech, but also music. The CELP algorithm is used for achieving a toll quality of speech at 13 kb/s. We have introduced a pitch predictor to restore the periodicity of speech signal. In order to reduce the computational complexity, we used an algebraic codebook and the Backward Filtering technique.