Comparison of frequency lowering algorithms on Mandarin speech recognition
Yunyi Zhang, Jie Cui, Ling Xiao · The Journal of the Acoustical Society of America · 2012
This paper examined several frequency lowering algorithms for improving speech recognition in Chinese listeners with high-frequency sensorineural hearing loss, including both frequency compression and frequency transposition algorithms. The frequency lowering algorithms were evaluated by a subjective listening test. Monosyllabic words in Mandarin spoken by a female and a male talker were used as speech materials. Speech materials processed by frequency compression and frequency transposition were presented to normal-hearing listeners in the listening tests, as well as the control condition of unprocessed speech. More benefits in speech recognition were observed with speech spoken by female talker than by male talker. It was showed that frequency lowering made little difference for some words while words with dental consonant were most infected by frequency lowering. The results indicate that Mandarin recognition for high-frequency hearing loss may benefit from frequency lowering in some cases, while whether choosing frequency lowering or not and which algorithm to choose depends on individuals.