Speech audiometry, as compared to pure tone audiometry, is of special importance since it directly evaluates subject's ability of speech perception, and has broadened the area for the differential diagnosis of retrocochlear deafness. Natural speech sounds used in conventional speech audiometry, however, impose serious limitations upon the reliability of results because of the variability in human utterances and the difficulty in the exact specification of their acoustic characteristics. On the other hand, synthetic speech sounds are ideally suited for this purpose since their acoustic characteristics can be exactly specified and easily controlled. Identification tests were performed on 100 cases of sensorineural hearing impairment using synthetic vowels. Analysis of response patterns in the F
1-F
2 plane leads to a classification of these cases into 4 types. The relationship between these types and thepure tone audiograms are examined, which indicates close correlation of response patterns and audiograms, in cases of Type 1 and Type 2.
1. Introduction
Although the pure tone audiogram provides an accurate means of investigating the subject's absolute threshold of hearing, it does not provide enough information as to the ability of speech perception in hearing-impaired patients. Speech audiometry, on the other hand, is intended to evaluate directly the subject's ability of speech perception, and has broadened the area of differential diagnosis of sensorineural deafness.
Natural speech sounds used in conventional speech audiometry, however, impose serious limitations upon the quantitative interpretation of the results, because of the variability in human utterances and the difficulty in exactly specifying their acoustic properties. These difficulties can he overcome by using synthetic speech sounds as test stimuli, since their acoustic characteristics can be exactly specified and easily controlled, so that not only typical sounds, but also a whole continuum of speech sounds can be generatedd and the subject's performance can be investigated in detail.
The present study is an attempt toward utilization of synthetic speech sounds for the purpose of standardizing the procedures for the evaluation of sensorineural hearing impairments. Synthetic vowel stimuli with various combinations of the first and the second formant frequencies (F
1 and F
2) were generated by a terminal-analog speech synthesizer, and were used for speech audiometry. The results of identification tests on subjects with sensorineural hearing impairments as well as with normal hearing are analyzed. Cases of sensorineural deafness are shown to be classified into four types on the basis of their characteristic patterns of response in the F
1-F
2 plane. Correlation between these types and their pure tone audiograms are examined.
2. Experimental Procedures Stimuli
The synthetic vowels were produced by a terminal analog speech synthesizer. The fleck diagram is shown in Fig. 1. A train of shaped pulses was used as the glottal source, and was fed to a cascaded connection of circuits representing the higher pole correction, the radiation characteristic and four formant characteristics. The fundamental frequency was set at 140Hz. The first and the second formants were used as variable parameters, while the third and the fourth formants were held constant at 2, 700 and 3, 500Hz. A set of 64 synthetic vowel stimuli was chosen by the combination of F
1 and F
2 values as shown in Table 1, The duration of each test sound was 800msec, and the rise time constant for pitch and intensity was 20msec. These sounds were recorded on tape in random order and were presented monaurally to each of the subjects through an earphone in a sound-proof room. The intensity of sound stimuli was kept at the Most Comfortable Level for each subject, which turned out to be between 80 to 100dB in sensation
View full abstract