[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
HSR results for Aurora2 data
does anyone know about human speech recognition (HSR) experiments performed with the Aurora2 speech data? Leonard reported an HSR score of over 99% for the clean TIDigits, which are being used for Aurora2. I'd be interested in human scores in the presence of the modulated noises in Aurora2, and compare those to automatic speech recognition results. I imagine that in the low-SNR conditions listeners actually produce some errors, but I couldn't find any publication on this.
Bernd T. Meyer
Carl von Ossietzky Universität Oldenburg
D-26111 Oldenburg, Germany