Hopfield Model Applied to Vowel and Consonant Discrimination
Generate an AI Snapshot to get a quick, structured summary of this paper.
A concise AI-generated summary of the paper will appear here once you click Generate AI Snapshot.
Abstract
A Hopfield model of 120 ‘‘neurons’’ has been simulated on an LDSP (Lincoln Digital Signal Processor). The model has been applied to the study of problems in automatic discrimination of vowels and consonants. In the first problem, a spectral cross section was extracted by performing a fast Fourier transform on a 20 ms segment from the steady state portion of the vowel in a single syllable word. The spectrum was then smoothed and a one‐bit gradient measure applied at 120 frequency values, thus obtaining an assigned state for that vowel. This procedure was repeated until eight such assigned states were obtained. From this data, the connection matrix Tij was obtained using the associative equation, Tij=Σ7s=0xsixsj Each xsi was the component of one of the assigned states.