杉山雅英
AUTOMATIC LANGUAGE RECOGNITION USING
ACOUSTIC FEATURES
Abstract:Language recognition (e.g. Japanese, English, German, etc) using acoustic features is an important
yet difficult problem for current speech technology. In this report, two language recognition algorithms
are proposed and some experimental results are described. The speech data base used in this report
contains 20 languages. The speech data was carefully divided into training and test sets, recognition
experiments being designed as both speaker-independent and text-independent. The first algorithm
is based on the standard Vector Quantization (VQ) technique. The second algorithm is based on a
single universal (common) VQ codebook for all languages, and its occurrence probability histograms.
The experimental results show that the recognition rates for the first and second algorithms were
65% and 80%, respectively, each using just 8 sentences of unknown speech (about 64 seconds). With
sufficient input speech the second algorithm is better than the first.