A　high　quality　speech　coding　algorithm　at　2kb/s　based　on　Waveform　interpolation　(WI)　is　presented　in　this　paper.　In　this　new　2kb/s　WI　speech　coding,　five　novel　techniques　were　incorporated,　which　are　Predictive　multi-stage　vector　quantization　(PMSVQ)　for　Line　spectral　frequency　(LSF),　pitch　estimation　algorithm　based　on　the　Dyadic　wavelet　transform　and　the　Normalized　cross-correlation　function　(DWT-NCCF),　predictive　Analysis-by-synthesis　(A-b-S)　vector　quantization　for　Slowly　evolving　waveform　(SEW)　magnitude　based　on　the　Discrete　cosine　transform　(DCT),　matrix　quantization　for　Rapidly　evolving　waveform　(REW)　magnitude　based　on　Variable　dimension　vector　quantization　(VDVQ)　and　DCT,　and　predictive　A-b-S　VQ　scheme　for　CW　power　based　on　the　temporal　weighting.　Subjective　quality　test　results　indicate　that　the　reconstructed　speech　quality　of　the　2kb/s　WI　speech　coding　exceeds　that　of　Mixed-excitation　linear　predictive　(MELP)　coding　at　2.4kb/s　greatly,　and　is　very　close　to　that　of　Code-excited　linear　predictive　(CELP)　coding　at　4.8kb/s.