Vowel recognition in noise for male, female and child voices

The effect of speaker adaptation in human recognition of speech in noise was investigated. A set of 15 vowels, in a /t/-vowel-/t/ context, was recorded for 5 male, 5 female, and 5 child speakers. Vowels were presented in a range of signal-to-noise ratios. Recognition scores were determined in three distinct conditions. In the first condition the material was presented per speaker. In the second, the speech items from speakers within one category (male, female, or child) were mixed. In the third, the items from all speakers were mixed. The three conditions were presented to groups of listeners in random order. The largest difference in threshold, found between the first and third condition, was only 2 dB