Waveform　interpolation　(WI)　speech　coding　has　a　good　performance　compared　with　various　low　bit　rate　coders,　and　provides　high-quality　reconstructed　speech.　However,　characteristic　waveform　(CW)　representation　and　alignment　in　WI　model　has　very　high　complexity　in　computation.　This　has　restricted　its　further　developments　and　practical　applications.　The　improved　algorithm　for　CW　representation　and　CW　alignment　was　proposed　by　using　FFT,　cubic　B-spline　interpolation　and　period　extension　in　this　paper.　The　experimental　results　show　that　the　improved　algorithm　has　greatly　reduced　the　computational　complexity　and　kept　the　same　speech　quality　as　original　coder.　©　2006　IEEE.