Perceptual evaluation of cost for segment selection in concatenative speech synthesis

In segment selection for concatenative text-to-speech (TTS), it is important to utilize a cost that corresponds to the perceptual characteristics. We clarify correspondence to the perceptual scores of the cost, and then various functions to integrate the costs are evaluated. The perceptual scores are determined from results of perceptual experiments on the naturalness of synthetic speech. The results show that the average cost, which shows the naturalness degradation over the entire synthetic speech has better correspondence to the perceptual scores than the maximum cost, which shows the local naturalness degradation. Furthermore, RMS (root mean square) cost, which is affected by both the average cost and the maximum cost, has the best correspondence.