Postprocessing techniques for voice pitch trackers

The concepts of dynamic programming and pattern matching are used in a postprocessing technique for voice pitch tracking. The dynamic programming is used to achieve a smooth pitch contour based on information in the current and previous frames. A voicing decision is made based on matching the smooth contour from the dynamic programming with a set of reference templates which are representative of voiced and unvoiced speech. Three postprocessing techniques; the dynamic programming technique, nonlinear smoothing and median smoothing are appended to three standard pitch tracking algorithms. A comparative evaluation of these postprocessing techniques is performed over a data base of 58 male and female speakers ranging from 6 to 87 years of age. It is shown that significant improvement in the performance of the pitch tracking algorithms can be obtained with the dynamic progranming postprocessing.