Synthetic Speech Discrimination using Pitch Pattern Statistics Derived from Image Analysis

摘要

In this paper, we extend the work by Ogihara, et al. to discrimi nate between human and synthetic speech using features based on pitch patterns. As previously demonstrated, significant dif ferences in pitch patterns between human and synthetic speech can be leveraged to classify speech as being human or synthetic in origin. We propose using mean pitch stability, mean pitch sta bility range, and jitter as features extracted after image analysis of pitch patterns. We have observed that for synthetic speech, these features lie in a small and distinct space as compared to human speech and have modeled them with a multivariate Gaussian distribution. Our classifier is trained using synthetic speech collected from the 2008 and 2011 Blizzard Challenge along with Festival pre-built voices and human speech from the NIST2002 corpus. We evaluate the classifier on a much larger corpus than previously studied using human speech from the Switchboard corpus, synthetic speech from the Resource Man agement corpus, and synthetic speech generated from Festival trained on the Wall Street Journal corpus. Results show 98% ac curacy in correctly classifying human speech and 96% accuracy in correctly classifying synthetic speech.