The Experts below are selected from a list of 45117 Experts worldwide ranked by ideXlab platform

Keiichi Tokuda - One of the best experts on this subject based on the ideXlab platform.

  • minimum generation error training by using Original Spectrum as reference for log spectral distortion measure
    International Conference on Acoustics Speech and Signal Processing, 2009
    Co-Authors: Keiichi Tokuda
    Abstract:

    This paper improves a minimum generation error (MGE) based HMM training technique for HMM-based speech synthesis by directly using the Original Spectrum instead of line spectral pairs (LSPs) as reference Spectrum for log spectral distortion (LSD) measure. Two types of Original reference spectra for LSD calculation are investigated, including the Spectrum extracted from speech waveform by STRAIGHT, and the short-time FFT Spectrum calculated from speech waveforms. Since only the harmonics of the FFT Spectrum are coincident with the underlying spectral envelope, the LSD between generated LSPs and Original FFT Spectrum is calculated by sampling at the harmonic frequencies, and a weighting function is designed to simulate the sampling strategy on LSPs. From the experimental results, the MGE-LSD training using the FFT Spectrum as reference Spectrum achieved the best performance.

  • ICASSP - Minimum generation error training by using Original Spectrum as reference for log spectral distortion measure
    2009 IEEE International Conference on Acoustics Speech and Signal Processing, 2009
    Co-Authors: Keiichi Tokuda
    Abstract:

    This paper improves a minimum generation error (MGE) based HMM training technique for HMM-based speech synthesis by directly using the Original Spectrum instead of line spectral pairs (LSPs) as reference Spectrum for log spectral distortion (LSD) measure. Two types of Original reference spectra for LSD calculation are investigated, including the Spectrum extracted from speech waveform by STRAIGHT, and the short-time FFT Spectrum calculated from speech waveforms. Since only the harmonics of the FFT Spectrum are coincident with the underlying spectral envelope, the LSD between generated LSPs and Original FFT Spectrum is calculated by sampling at the harmonic frequencies, and a weighting function is designed to simulate the sampling strategy on LSPs. From the experimental results, the MGE-LSD training using the FFT Spectrum as reference Spectrum achieved the best performance.

Josuke Okda - One of the best experts on this subject based on the ideXlab platform.

  • ICASSP - Analysis of a moving sound source-Orbit estimation using linear sensor array
    ICASSP '86. IEEE International Conference on Acoustics Speech and Signal Processing, 1
    Co-Authors: Seizi Nishimura, Tsuyoshi Usagawa, Masanao Ebata, Josuke Okda
    Abstract:

    It is not easy to observe the Original Spectrum information of a moving sound source because of the Doppler effect. In another paper presented in this proceeding, we propose the method to compensate the Doppler effect, which requires the orbit information of target sound source. This paper describes the orbit estimation method using three sensors. This method is composed of two parts; position estimation derived from a pair of relative propagation delays which is observed using cross-correlation technique, and orbit estimation by dynamical filtering using Kalman filter. Simple simulation was carried out to evaluate the performance of proposed method. The result shows the sufficient accuracy of estimation.

Seizi Nishimura - One of the best experts on this subject based on the ideXlab platform.

  • ICASSP - Analysis of a moving sound source-Orbit estimation using linear sensor array
    ICASSP '86. IEEE International Conference on Acoustics Speech and Signal Processing, 1
    Co-Authors: Seizi Nishimura, Tsuyoshi Usagawa, Masanao Ebata, Josuke Okda
    Abstract:

    It is not easy to observe the Original Spectrum information of a moving sound source because of the Doppler effect. In another paper presented in this proceeding, we propose the method to compensate the Doppler effect, which requires the orbit information of target sound source. This paper describes the orbit estimation method using three sensors. This method is composed of two parts; position estimation derived from a pair of relative propagation delays which is observed using cross-correlation technique, and orbit estimation by dynamical filtering using Kalman filter. Simple simulation was carried out to evaluate the performance of proposed method. The result shows the sufficient accuracy of estimation.

Tao Chen - One of the best experts on this subject based on the ideXlab platform.

  • Method for identifying transgenic cottons based on terahertz spectra and WLDA
    Optik, 2015
    Co-Authors: Jianjun Liu, Aijun Zhu, Fangrong Hu, Yong Du, Zhi Li, Tao Chen, Haitao Xin
    Abstract:

    Terahertz spectroscopy and weighted linear discriminant analysis (WLDA) are applied to identify different transgenic cottons in this paper. Principal component analysis (PCA) is applied to extract relevant features from Original Spectrum information and eliminate the anomalous samples. Instead of Original spectral information, the feature Spectrum is selected to fed into the model of WLDA, where an improved linear discrimination method to identify those samples. The results demonstrate that the WLDA method can effectively identify the distinct transgenic cottons, and terahertz spectroscopy can provide a nondestructive, rapid and reliable method to distinguish different transgenic cottons.

  • Identification of biomolecules by terahertz spectroscopy and fuzzy pattern recognition.
    Spectrochimica acta. Part A Molecular and biomolecular spectroscopy, 2013
    Co-Authors: Tao Chen
    Abstract:

    An approach for automatic identification of terahertz (THz) spectra of biomolecules is proposed based on principal component analysis (PCA) and fuzzy pattern recognition in this paper, and THz transmittance spectra of some typical amino acid and saccharide biomolecular samples are investigated to prove its feasibility. Firstly, PCA is applied to reduce the dimensionality of the Original Spectrum data and extract features of the data. Secondly, instead of the Original Spectrum variables, the selected principal component scores matrix is fed into the model of fuzzy pattern recognition, where a principle of fuzzy closeness based optimization is employed to identify those samples. Results demonstrate that THz spectroscopy combined with PCA and fuzzy pattern recognition can be efficiently utilized for automatic identification of biomolecules. The proposed approach provides a new effective method in the detection and identification of biomolecules using THz spectroscopy.

T. Parks - One of the best experts on this subject based on the ideXlab platform.

  • ICASSP - Deterministic estimation of two-dimensional signals
    ICASSP '83. IEEE International Conference on Acoustics Speech and Signal Processing, 1
    Co-Authors: Sergio D. Cabrera, T. Parks
    Abstract:

    A technique for spectral estimation given a finite number of measurements of a two-dimensional band-limited signal of time t and space x is used on signals which are transient propagating waves, The theory of optimal estimation for deterministic signals is presented in two-dimensions for signals which are assumed to be contained in a wedge-shaped region of the two-dimensional frequency plane. We also consider samples which contain errors. An example with exact signal samples shows a spectral estimate with all the basic features of our test signal. After adding noise to this signal, we encounter more difficulties in recovering the Original Spectrum.