H.263 Facial Video Image Coding with the MELP Voice Codec

The International Telecommunication Union has put forward a recommendation H.263 “Video Coding for Low Bitrate Communication”. This recommendation defines the rules for compressing the moving picture component of audio-visual services at low bit rates. The H.263 video codec developed from MPEG-1 is primarily designed for low bit rate applications. However, there is still room to improve its encoding speed if its application is limited to videophone and videoconferencing which uses the internet or ISDN phone lines. The basic encoding operations involved in motion compensation, discrete cosine transform (DCT) and bit rate control were modified and tailored specifically for facial video images. The strategies to improve the encoding speed are presented in this paper. Experimentally, we achieved an ecoding speed approximately 10 times faster than that of the generic H.263. While imposing a restriction on the video object bing facial images, a low bit rate sound layer based on the MELP vocoder was added by sparing 2.4 kbps for voice communication as a part of H.263 video codec. Fluctuations both in terms of the waveform and pitch period of voiced speech was incorporated in the MEPL vocoder to enhance the naturalness of sysnthsized speech.