에너지 연산자에 기초한 간단한 피치 추적 방법

A Simple Pitch Tracking Algorithm based on the Energy Operator

  • 발행 : 2004.01.01

초록

유성음의 피치주파수 궤적을 추정할 수 있는 새로운 방법을 제시하였다. 이 방법은 에너지연산자[1]를 두 번 적용하는데 기초하고 있다. Kaiser의 에너지연산자는 정현파의 진폭과 주파수 정보를 추출하는 기능을 가지고 있다. 변조모형에 의하면 유성음은 피치 신호로 변조된 포만트들의 합성으로 파악될 수 있으므로 이 파형의 진폭 포락선을 추출해서 피치 신호와 유사한 파형을 얻는다. 이 파형의 평균 주파수를 검출하여 피치 주파수를 구하는 것이다. 앞부분은 Gopalan의 접근법[9]과 마찬가지이나, 뒷부분의 LPC-스펙트럼 분석등의 과정 대신 또 한번 에너지 연산자를 적용하도록 하여 매우 단순화되고 온라인 적용이 가능한 알고리듬을 얻었다. 추정 결과는 거친 편이지만 온라인으로 피치 궤적의 일반적 스케치를 얻는데 유용할 것으로 기대된다.

A new method for the estimation of pitch-frequency contour of voiced speech is presented. The method is based on the double application of Kaiser's energy operator[1], which has the capabilities of extracting amplitude and frequency of a sinusoidal waveform. According to the modulation model, a vowel can be represented by a combination of damped sinusoids representing formants, modulated by pitch pulses. Therefore, the amplitude envelope of each of the components will give a pitch-like waveform and the pitch can be obtained by averaging the frequencies of this waveform. The first part is the same as Gopalan's approach[9], but by substituting the LPC based spectral analysis with the second application of energy operator, the algorithm becomes very simple and can be processed on-line. Although the estimation is rather coarse, the suggested algorithm can be useful for getting a general sketch of pitch contour on-line.

키워드

참고문헌

  1. Proc. IEEE ICASSP 90 On a simple algorithm to calculate the 'energy' of a signal J.F.Kaiser
  2. Proc. IEEE ICASSP 92 On separating amplitude from frequency modulations using energy opergy operators P.Maragos;J.F.Kaiser;T.F.Quatieri
  3. IEEE Trans. on Signal Processing v.42 no.2 Conditions for positivity of an energy operator A.C.Bovik;P.Maragos
  4. Proc. IEEE ICASSP 01 An improved demodultion algorithm using splines D.Dimitriadis;P.Maragos
  5. IEEE Trans. on Signal Processing v.41 no.10 Energy separation in signal modulations with application to speech analysis P.Magros;J.F.Kaiser;T.F.Quatieri
  6. Proc. IEEE ICASSP 91 Speeh nonlinearity, modulation and energy operators P.Magros;J.F.Kaiser;T.F.Quatieri
  7. IEEE Signal Processing Letters v.1 no.11 Energy onset times for speaker identification T.F.Quatieri;C.R.Jankowski;Jr.;D.A.Reynolds
  8. Record of 28th Asilomar Conference on Sinals, Systems and Computers An investigation of estimating pitch period using a non-linear differential operator R.K.Whitman;D.M.Etter
  9. Proc. WCCC-ICSP2000 Pitch estimation using a modulation model of speech K.Gopalan
  10. Proc. IEEE ICASSP 92 Application of the modulation model to speech recognition A.B.Finberg;R.J.Mammone;J.L.Flanagan
  11. Speech Processing and Synthesis Toolboxes D.G.Childers
  12. IEEE Trans. on Signal Processing v.41 AM-FM energy detection and separation in noise using multiband energy operators A.C.Bovik;P.Maragos;T.F.Quatieri
  13. Proc. IEEE ICASSP 84 v.9 A performance comparison of pitch extraction algorithms for noisy speeh K.A.Oh;C.K.Un
  14. IEEE Trans. on Aucoustics, Speech, and Signal Processing v.ASSP-26 no.4 Real-time harmonic pitch detector S.Seneff
  15. Proc. IEEE ICASSP 90 v.1 Pitch estimation and voicing detection based on a sinusoidal speech model R.J.McAulay;T.F.Quatieri