Title :
Vowel synthesis by on-line EM algorithm with IIR filter
Author :
Tamakoshi, Hiroki ; Ishii, Shin ; Yoshida, Wako ; Sato, Masa-aki
Author_Institution :
Nara Inst. of Sci. & Technol., Japan
Abstract :
We formerly proposed an online EM algorithm for normalized Gaussian network, and showed its capability to learn and reconstruct chaotic dynamical systems. Using the integral coordinate embedding method with IIR filters, this method is robust to noise. In this article, we apply the method to learning of waveforms of Japanese vowels. Experimental results show the learning and the synthesis of vowels is successful. In addition, irregular fluctuations of waveforms, which are important for their naturalness, are also preserved
Keywords :
Gaussian processes; IIR filters; chaos; filtering theory; learning (artificial intelligence); maximum likelihood estimation; neural nets; online operation; signal reconstruction; speech synthesis; IIR filter; Japanese vowel waveforms; chaotic dynamical system learning; chaotic dynamical systems reconstruction; integral coordinate embedding method; irregular fluctuations; normalized Gaussian network; online EM algorithm; vowel learning; vowel synthesis; Chaos; Covariance matrix; Fluctuations; IIR filters; Integrated circuit modeling; Linear regression; Noise robustness; Probability; Stochastic processes; Vectors;
Conference_Titel :
Neural Networks, 2001. Proceedings. IJCNN '01. International Joint Conference on
Conference_Location :
Washington, DC
Print_ISBN :
0-7803-7044-9
DOI :
10.1109/IJCNN.2001.938823
