Abstract: A scalable variable rate characteristic waveform interpolation (CWI) speech coding algorithm is proposed based on wavelet transform (WT). The 2-D characteristic waveform surface is decomposed into several spaces with different time-frequency resolution, which serves as the enhancement layers of scalable CWI coding scheme. The basic layer only contains linear prediction parameters, pitch cycle and power normalized factors. The scalable WT-CWI coder obtains variable bit rate. MUSHRA subjective listening test showed that with the coding rate increasing from 1.8kbit/s to 3.6kbit/s, the basic layer of the proposed scalable variable rate CWI coder outputs acceptable speech quality and the enhancement layers improved the speech quality in turn.