Development and Testing of a Formant‐Coding Speech Compression System

The development of a speech‐band‐width compression system employing formant‐coding principles is described. The compression system codes the input speech in terms of seven electrical signals, representing the frequencies of the first three formants, the amplitudes of voicing and of friction, the fundamental vocal frequency, and the frequency of the spectral maximum of the fricative excitation. These signals occupy a total band width of the order of 60 cps and require signal‐to‐noise ratios of approximately 30 db for their transmission.An evaluation of the intelligibility of speech transmitted by the compression system is described. Natural monosyllabic utterances are transmitted through the system and presented to listeners for identification. Vowel and consonant articulation scores are computed and sound confusion matrices are constructed and discussed. The results indicate that correct identification occurred for approximately 80% of the vowels and 25% of the consonants.