Höge, H. (Siemens AG, Corporate Technology)
Kotnik, B.; Kacic, Z. (University of Maribor)
Pfitzinger, H. R. (Ludwig-Maximilians-Universität München)
Epoch detection plays an increasing role in speech processing. For this purpose pitch marking algorithms (PMAs) working accurately in various environments are needed. To push development of PMAs a suited framework for evaluation is needed. In this paper criteria describing the performance of PMAs are proposed and two databases which allow to evaluate the PMAs, are presented. On the basis of a new PMA the proposed framework for evaluation is tested. The PMA achieved its highest success rate of 95.2 % with a close-talk channel.