ML Parameter Generation with a Reformulated MGE Training Criterion — Participation in the Voice Conversion Challenge 2016

Abstract

This paper describes our entry to the Voice Conversion Challenge 2016. Based on the maximum likelihood parameter generation algorithm, the method is a reformulation of the minimum generation error training criterion. It uses a GMM for soft classification, a Mel-cepstral vocoder for acoustic analysis and an improved dynamic time warping procedure for source-target alignment. To compensate the oversmoothing effect, the generated parameters are filtered through a speaker-independent postfilter implemented as a linear transform in cepstral domain. The process is completed with mean and variance adaptation of the log-fundamental frequency and duration modification by a constant factor. The results of the evaluation show that the proposed system achieves a high conversion accuracy in comparison with other systems, while its naturalness scores are intermediate.

LSTM based voice conversion for laryngectomees

“…Due to the characteristics of the oesophageal speech, there is a very important mismatch between both healthy and oesophageal signals which causes the inadequacy of using a dynamic time warping (DTW) algorithm directly [31]. This is why both signals were labelled at phone level, and then the iterative alignment procedure described in [32] and [33] was applied for each pair of oesophageal and healthy phones.…”