Pattern identification by analysis of digital words

Abstract

A method of and an apparatus for analyzing multi-bit digital words to identify their source or content. The digital words are converted to multi-bit fields, with each field being a portion of the digital word or corresponding to one or more descriptive parameters of the source. Each multi-bit field is analyzed to generate an additional multi-bit field. The additional multi-bit fields are grouped into pairs, and each pair is analyzed to generate a further multi-bit field. The further multi-bit fields are grouped into pairs. If more than one pair exists, the analyzing and grouping into pairs continues until a single pair remains, and that pair is analyzed to identify the source or content. Preferably each of the additional and further fields has the same number of bits, and each of the means for analyzing the additional and further multi-bit fields has twice that number of input bits. If the means for analyzing multi-bit fields has enough input bits, the additional and further fields could be grouped into groups larger than pairs. The analysis of the original fields can be done by comparison of each field with known bit patterns associated with known sources to determine possible sources, or by the application of the fields as the address bits to a programmable memory to select a memory location containing a source number or ID code indicative of possible sources. The subsequent analysis of pairs of fields can be the application of the pairs of fields as the address bits to a programmable memory to select a memory location containing a source number or ID code indicative of common possible sources. Each field can represent more than one parameter, and the analysis of the original fields can look for sources whose parameters closely but not exactly match those in the analyzed fields. The degree and measure of closeness required is selectable during programming of the pattern identifier. The degree of closeness need not be fixed, and the analysis of the original fields can look for the one known source whose parameters most closely match those in the analyzed fields. The method can be carried out by a digital data processor.