Moment-based fast discrete sine transforms

This paper presents a novel approach to compute discrete sine transforms (DSTs). By using a modular mapping, DSTs are approximated by the sum of a finite sequence of discrete moments. Hence, by extending our earlier technique in computing moments with an adder network only, DSTs can also be implemented easily by a systolic array primarily involving additions. The method can be applied to multidimensional DSTs as well as their inverses.