Entropy, and the recognition of fuzzy letters

Amnon Meisels, Abraham Kandel, Gay Gecht

Research output: Contribution to journalArticlepeer-review

1 Scopus citations


The behavior of information in a process of recognition is of much theoretical interest. In order to shed some light on this behavior a case study is presented. The process chosen was that of the recognition of several computer printed characters on a noisy page. The design of the experiment was made such that the amount of missing information in the data was exactly calculable. The recognition process is iterative and is found to always terminate. Another result is that information is gained, on the average, during the process of recognition. This was true for all different characters that participated in the analysis. A connection with the theory of fuzzy pattern recognition is very clear and is discussed in detail. A correlation was found to exist between either the final average entropy per symbol, or the drop in entropy during the iterative process, and the percentage of successful detection of the same specific symbol. This fact further connects missing information with the process of learning and recognizing fuzzy data.

Original languageEnglish
Pages (from-to)297-309
Number of pages13
JournalFuzzy Sets and Systems
Issue number3
StatePublished - 20 Jul 1989


  • Pattern recognition
  • detection
  • entropy
  • fuzzy data

ASJC Scopus subject areas

  • Logic
  • Artificial Intelligence


Dive into the research topics of 'Entropy, and the recognition of fuzzy letters'. Together they form a unique fingerprint.

Cite this