The Journal of the Acoustical Society of America · 1964 · 14 citations · 0 references
Language StudiesLinguisticsAcousticsPhoneticsPsychoacousticsAcoustic PhoneticsAuditory DiscriminationAuditory PhoneticsSpeech AcousticsSpeech RecognitionSpeech Sound DiscriminationAuditory Signal DetectionFrequency DiscriminationFrequency Spectrum AnalysisPeriodicityPeriodicity DetectionPhonetic MeasurementSpeech InputSpeech Signal Analysis
Speech sounds can be approximately divided into two categories, buzz or periodic sounds and hiss or aperiodic sounds. Vocoder buzz-hiss detectors have discriminated between the two classes by comparing the high-frequency and low-frequency portion of the spectrum. This technique succeeds during most of the input speech because most voiceless sounds contain a preponderance of high-frequency energy whereas the voiced sounds emphasize the lows. However, voiceless sounds such as aspirates and plosives often contain large amounts of low-frequency energy. Also, microphone distortions can produce low-frequency aperiodic sounds. This paper describes a periodicity meter that processes the output from six pitch detectors to produce another measure appropriate for buzz-hiss detection. Preliminary results are presented.