초록 ▼

A signal coding technique for use with fractional bit rates is disclosed which, for a given dimensionality, allows the use of smaller constellations, and thus provides greater robustness in the presence of various impairments, than would otherwise be the case yet entails essentially no decoding dela

A signal coding technique for use with fractional bit rates is disclosed which, for a given dimensionality, allows the use of smaller constellations, and thus provides greater robustness in the presence of various impairments, than would otherwise be the case yet entails essentially no decoding delay penalty. In this technique, one or more of the input bits are trellis encoded and the trellis-encoded bits identify one of a predetermined plurality of subsets of the signal constellation and the particular symbols selected for transmission from two or more identified subsets are chosen by the non-trellis-encoded bits interdependently.

대표청구항 ▼

A method for encoding, into 2N-dimensional symbols, N being an integer, a stream of input bits occurring at a fractional bit rate of r bits per 2N-dimensional symbol relative to a predetermined symbol rate, said 2N-dimensional symbols being selected from a 2N-dimensional constellation comprised of a

A method for encoding, into 2N-dimensional symbols, N being an integer, a stream of input bits occurring at a fractional bit rate of r bits per 2N-dimensional symbol relative to a predetermined symbol rate, said 2N-dimensional symbols being selected from a 2N-dimensional constellation comprised of a concatenation of N/M 2M-dimensional constellations each of which is partitioned into a plurality of 2M-dimensional subsets of signal points, the method comprising the steps of encoding first ones of the input bits to identify a sequence of 2M-dimensional subsets, and identifying signal points from successive ones of the identified 2M-dimensional subsets jointly in response to at least a particular group of the others of the input bits, said particular bit group being associated with at least L 2N-dimensional symbol intervals, L being an integer greater than unity.