CONF Marelli_ICASSP2019_2019/IDIAP An End-to-end Network to Synthesize Intonation Using a Generalized Command Response Model Marelli, François Schnell, Bastian Bourlard, Hervé Dutoit, T. Garner, Philip N. Digital IIR Filters Fujisaki Model neural networks Prosody Modelling speech synthesis EXTERNAL https://publications.idiap.ch/attachments/papers/2019/Marelli_ICASSP2019_2019.pdf PUBLIC https://publications.idiap.ch/index.php/publications/showcite/Marelli_Idiap-RR-05-2019 Related documents ICASSP 2019 - 2019 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP) Brighton, United Kingdom 2019 IEEE 7040-7044 https://ieeexplore.ieee.org/document/8683815 URL 10.1109/ICASSP.2019.8683815 doi The generalized command response (GCR) model represents intonation as a superposition of muscle responses to spike command signals. We have previously shown that the spikes can be predicted by a two-stage system, consisting of a recurrent neural network and a post-processing procedure, but the responses themselves were fixed dictionary atoms. We propose an end-to-end neural architecture that replaces the dictionary atoms with trainable second-order recurrent elements analogous to recursive filters. We demonstrate gradient stability under modest conditions, and show that the system can be trained by imposing temporal sparsity constraints. Subjective listening tests demonstrate that the system can synthesize intonation with high naturalness, comparable to state-of-the-art acoustic models, and retains the physiological plausibility of the GCR model. REPORT Marelli_Idiap-RR-05-2019/IDIAP AN END-TO-END NETWORK TO SYNTHESIZE INTONATION USING A GENERALIZED COMMAND RESPONSE MODEL Marelli, François Schnell, Bastian Bourlard, Hervé Dutoit, T. Garner, Philip N. Digital IIR Filters Fujisaki Model neural networks Prosody Modelling speech synthesis EXTERNAL https://publications.idiap.ch/attachments/reports/2018/Marelli_Idiap-RR-05-2019.pdf PUBLIC Idiap-RR-05-2019 2019 Idiap May 2019 The generalized command response (GCR) model represents intonation as a superposition of muscle responses to spike command signals. We have previously shown that the spikes can be predicted by a two-stage system, consisting of a recurrent neural network and a post-processing procedure, but the responses themselves were fixed dictionary atoms. We propose an end-to-end neural architecture that replaces the dictionary atoms with trainable second-order recurrent elements analogous to recursive filters. We demonstrate gradient stability under modest conditions, and show that the system can be trained by imposing temporal sparsity constraints. Subjective listening tests demonstrate that the system can synthesize intonation with high naturalness, comparable to state-of-the-art acoustic models, and retains the physiological plausibility of the GCR model.