Speech/non-speech detection in Malay language spontaneous speech

摘要: The goal of this work is to discriminate speech and non-speech segments in Malay language spontaneous as speech/non-speech detection important many processing applications. Inaccurate sentence boundaries are a major cause errors automatic recognition preprocessing stage that the signal into periods invaluable improving accuracy. We proposed combination three audio features energy, zero crossing rate (ZCR) fundamental frequency (F0) for each feature has unique properties differentiate segments. Experiments conducted on one-hour consisting more than 20,000 An accuracy evaluation reveals method achieved 97.8% rate. Non-speech will further be used candidates boundary our next experiment.