Wu-Ton Chen

Deconvolution and vocal-tract parameter estimation of speech signals by higher-order statistics based inverse filters

The authors propose a two-step method for deconvolution and vocal-tract parameter estimation of (non-Gaussian) voiced speech signals. In the first step, the driving input (a non-Gaussian pseudo-periodic positive pulse train) to the vocal-tract filter which can be nonminimum-phase is estimated from speech data by a higher-order statistics (HOS) based inverse filter. In the second step, autoregressive moving average (ARMA) parameters of the vocal-tract filter are estimated with the estimated input and speech data by a prediction error system identification method (an input-output system identification method). Finally, some experimental results with real speech data are provided.<<ETX>>

New cumulant-based inverse filter criteria for deconvolution of nonminimum phase systems

This work proposes a new family of cumulant-based inverse filter criteria J/sub M,m/, which require a single slice of Mth-order (M/spl ges/3) cumulants, a (2m)th-order cumulant, and a (2M-2m)th-order cumulant of the inverse filter output where 1/spl les/m/spl les/M-1, for deconvolution of linear time invariant (LTI) nonminimum phase systems with only non-Gaussian output measurements contaminated by Gaussian noise. Some simulation results are then presented for a performance comparison of the proposed criteria, Tugnaits (1993) criteria, and Chi and Kungs (1992) criteria. Finally, conclusions are presented.

An adaptive maximum-likelihood deconvolution algorithm

Abstract Kormylo and Mendel proposed a maximum-likelihood deconvolution (MLD) algorithm for estimating a desired sparse spike sequence μ ( k ), modelled as a Bernoulli-Gaussian (BG) sigbal, which was distorted by a linear time-invariant system v ( k ). Then Chi, Mendel and Hampson proposed another MLD algorithm which is a computationally fast MLD algorithm and has been successfully used to process real seismic data. In this paper, we propose an adaptive MLD algorithm, which allows v ( k ) to be a slowly time-varying linear system, for estimating the BG signal μ ( k ) from noisy data. Like the previous MLD algorithms, the proposed adaptive MLD algorithm can also recover the phase of v ( k ) when v ( k ) is time-invariant. Some simulation results are provided to support the proposed algorithm.

New inverse filter criteria for identification and deconvolution of nonminimum-phase systems by single cumulant slice

The authors propose a family of new cumulant based inverse filter criteria which only require a single slice of cumulants of the inverse filter output for the identification and deconvolution of linear time-invariant (LTI) nonminimum-phase systems with only non-Gaussian output measurements contaminated by Gaussian noise. Some simulation results and an application to speech deconvolution are provided to demonstrate that inverse filtering algorithms based on the proposed new criteria work well.<<ETX>>

Linear prediction based on higher order statistics by a new criterion

This criterion requires only partial Mth-order cumulants C/sub M,e/(0,k/sub 1/, k/sub 1/, . . ., k/sub M/2-1/, k/sub M/ /sub /2-1/) of the prediction error e(k) where M is even. Theoretically, it is shown that the proposed filter associated with a stationary process x(k) is the same as the conventional correlation based (minimum-phase) LPE filter associated with the nonGaussian signal y(k) (noise-free). Simulation results show that when y(k) is an autoregressive process of known order, the proposed filter works well.<<ETX>>

Chattering alleviation of variable structure control for a certain class of nonlinear systems

Abstract A new variable structure control algorithm is developed for a certain class of nonlinear systems. We have shown that the serious problem of chattering about the switching surfaces of variable structure systems can be alleviated by this algorithm. To illustrate its design procedures, simulation of a two‐link nonlinear robotic manipulator is considered. The simulation results indicate that benefit is not obtained at the expense of time required to reach steady‐state.

An adaptive Bernoulli-Gaussian model based maximum-likelihood channel equalizer for detection of binary sequences

Based on a modified Bernoulli-Gaussian model, an adaptive maximum-likelihood channel equalizer is proposed. It is a block signal processing algorithm for the detection of binary sequences transmitted through an unknown slowly time-varying channel. Both computational load and storage required by the proposed adaptive channel equalizer are linearly rather than exponentially proportional to the size of signal processing block. A simulation example is provided to show that it can simultaneously track the variation of slowly time-varying channels and detect unknown binary sequences well.<<ETX>>

Maximum-Likelihood Blind Deconvolution: Non-White Bernoulli-Gaussian Case

Todoeschuck and Jensen (1-21 recently reported that some reflectivity sequences p(k) calculated from sonic logs are not white and have a power spectral density approximately proportional to frequency, called a Joseph spectrum. The well-known MLD algorithms 7 131 can simultanmusly provide estimates of p(k), source wavelet wkh need not be minimum-phase, and statistical parameters. Although these MLD algorithms work well, they are based on the white Bernoulli-Gaussian (B-G) model for p(k). In this paper, assuming that spectrum measurements of p(k) are available, we propose a ML algorithm for blind deconvolution as p(k) is non-white with a general spectrum mywhile the spectrum of the obtained maximum-iikelihood estimate %L(k) is consistent with the measured spectrum.