David Labarre

Dual

This paper deals with the joint signal and parameter estimation for linear state-space models. An efficient solution to this problem can be obtained by using a recursive instrumental variable technique based on two dual Kalman filters. In that case, the driving process and the observation noise in the state-space representation for each filter must be white with known variances. These conditions, however, are too strong to be always satisfied in real cases. To relax them, we propose a new approach based on two dual Hinfin filters. Once a new observation of the disturbed signal is available, the first Hinfin algorithm uses the latest estimated parameters to estimate the signal, while the second Hinfin algorithm uses the estimated signal to update the parameters. In addition, as the Hinfin filter behavior depends on the choice of various weights, we present a way to recursively tune them. This approach is then studied in the following cases: (1) consistent estimation of the AR parameters from noisy observations and (2) speech enhancement, where no a priori model of the additive noise is required for the proposed approach. In each case, a comparative study with existing methods is carried out to analyze the relevance of our solution.

{H}_{\infty}

The estimation of the parameters of an autoregressive process (AR) from noisy observations is still a challenging problem. In this paper, we propose to sequentially estimate both the signal and the parameters, avoiding a non-linear approach such as the extended Kalman filter. The method is based on two conditionally linked Kalman filters running in parallel. Once a new observation is available, the first filter uses the latest estimated AR parameters to estimate the signal, while the second filter uses the estimated signal to update the AR parameters. This approach can be viewed as a recursive instrumental variable-based method and hence has the advantage of providing consistent estimates of the parameters from noisy observations. A comparative study with existing algorithms illustrates the performances of the proposed method when the additive noise is either white or coloured.

Algorithms for Signal Processing— Application to Speech Enhancement

When a single sequence of noisy observations is available, the autoregressive (AR)-model based methods using Kalman-filter make it possible to enhance speech. However, the estimation of the AR parameters is required, but is still a challenging problem as the signal is corrupted by an additive noise. In this paper, we propose to both estimate the signal and the AR parameters by developing a recursive instrumental variable-based approach. Avoiding a non linear approach such as the EKF, this method involves two conditionally linked Kalman filters running in parallel. Once a new observation is available, the first filter uses the latest estimated AR parameters to estimate the signal, while the second filter uses the estimated signal to update the AR parameters. A comparative study between existing speech enhancement methods is completed.

Consistent estimation of autoregressive parameters from noisy observations based on two interacting Kalman filters

Among parametric methods for speech enhancement, one consists in combining an autoregressive model for speech and a Kalman filter. This filtering is optimal in the H/sub 2/ sense providing the initial state vector, the input and the observation vectors in the state space representation of the system are independent, white and Gaussian. However, these assumptions do not necessarily hold when processing speech. In this paper, we propose to investigate an alternative approach, which is based on H/sub /spl infin// filtering and hence does not depend on these restrictive assumptions. In that setting, the purpose is to minimize the worst possible effects of the noises and system uncertainties on the estimation error. A comparative study between Kalman and H/sub /spl infin// filtering is carried out, when the additive colored noise can be modeled by a moving average (MA) process.

Two-Kalman filters based instrumental variable techniques for speech enhancement

In this paper, we propose a new Rayleigh channel simulator. Modeling the channel by an AR process leads to numerical problems due to the bandlimitation of the theoretical power density spectrum (PSD) of a Rayleigh channel. Therefore, we suggest modeling the channel by a low-pass filtered version of the so-called stochastic sinusoidal process. It consists of sinusoids in quadrature with random magnitudes modeled as AR processes. To estimate the AR parameters of the amplitudes, we take advantage of the asymptotic behavior of the first-kind zero-order Bessel function. We show that unlike an AR channel modeling, this simulator has the advantage of exhibiting the PSD peaks at the maximum Doppler frequency, for any AR process order.

Relevance of H ∞ filtering for speech enhancement

Among parametric methods for speech enhancement, one consists in combining an autoregressive model for speech and a Kalman filter. This filtering is optimal in the H/sub 2/ sense providing the initial state vector, the input and the observation vectors in the state space representation of the system are independent, white and Gaussian. However, these assumptions do not necessarily hold when processing speech. In this paper, we propose to investigate an alternative approach, which is based on H/sub /spl infin// filtering and hence does not depend on these restrictive assumptions. In that setting, the purpose is to minimize the worst possible effects of the noises and system uncertainties on the estimation error. A comparative study between Kalman and H/sub /spl infin// filtering is carried out, when the additive colored noise can be modeled by a moving average (MA) process.

The Stochastic Sinusoidal Model for Rayleigh Fading Channel Simulation

Among parametric methods for speech enhancement, one consists in combining an autoregressive model for speech and a Kalman filter. This filtering is optimal in the H/sub 2/ sense providing the initial state vector, the input and the observation vectors in the state space representation of the system are independent, white and Gaussian. However, these assumptions do not necessarily hold when processing speech. In this paper, we propose to investigate an alternative approach, which is based on H/sub /spl infin// filtering and hence does not depend on these restrictive assumptions. In that setting, the purpose is to minimize the worst possible effects of the noises and system uncertainties on the estimation error. A comparative study between Kalman and H/sub /spl infin// filtering is carried out, when the additive colored noise can be modeled by a moving average (MA) process.