On the computation of running discrete cosine and sine transform

Two algorithms are given for the computation of the updated discrete cosine transform-II (DCT-II), discrete sine transform-II (DST-II), discrete cosine transform-IV (DCT-IV), and discrete sine transform-IV (DST-IV). It is pointed out that the algorithm used for running DCT-IV can also be used for computation for running DST-IV without additional computational overhead. An architecture which is common and suitable for VLSI implementation of the derived algorithms is also presented. Preliminary studies have shown that the architecture can easily be implemented in VLSI form, and, in conjunction with a high-speed digital signal processor (for example ADSP 2100A), it can be used for real-time transform domain LMS adaptive filtering (128 taps) of 8 kHz sample rate speech signals.<
>