The detection of pitch accent is significant for both speech recognition and language understanding since the pitch accented words usually contain new and important information. This paper presents a novel feature called spectral balance-based cepstra...
The detection of pitch accent is significant for both speech recognition and language understanding since the pitch accented words usually contain new and important information. This paper presents a novel feature called spectral balance-based cepstral coefficient (SBCC) to capture the spectral characteristic of pitch accentuation. We used the Boston Radio News Corpus (BRNC) for experiments on the speaker independent pitch accent detection. The experimental results showed that the automatic labeling of pitch accent exhibited an average of 81.2% agreement with the hand labels.