Signal processing to improve speech intelligibility in perceptive deafness.

Abstract

A deaf person with recruitment perceives sound as though listening through a volume expander followed by an attenuator, the expansion ratio and attenuation being typically frequency dependent. (Other perceptive aberrations may also be present, of course.) The subject is often prevented from using enough hearing-aid gain to bring weak consonants into the useful dynamic range of his hearing, because this amount of gain would make lower-frequency, high-amplitude vowels intolerably loud. Such subjects commonly find amplified speech to have poor intelligibility. In a preliminary experiment it is established that recruitment in normal subjects, induced by masking or simulated by expansion of the signal, reduces the intelligibility of amplified speech severely, and that this intelligibility can be largely restored by signal processing. The implication is that recruitment in deaf subjects is a sufficient cause for loss of intelligibility, whether or not there are other causes. In the present experiments, speech is processed by a two-channel amplitude compressor whose frequency-dependent compression ratio is adjusted to compensate the recruitment of the individual subject, and the compressed speech is subjected to frequency-selective amplification similarly adapted to the subject. The aim is to amplify each acoustical element of speech, at each frequency-amplitude coordinate of the speech band, to a relative loudness for the deaf subject corresponding to the relative loudness of that speech element perceived by normals. This processing improved speech recognition, both in quiet and in the presence of competing speech introduced before processing, for six perceptively deaf subjects. Subjects showed an improvement in either initial- or terminal-consonant recognition of at least 22% and as much as 160% at optimum levels in quiet, and from 10% to 229% with speech interference 10 dB below the pre-processed signal.