This paper proposes a modular and flexible approach to adaptive Kalman filtering using the framework of a mixture-of-experts regulated by a gating network. Each expert is a Kalman filter modeled with a different realization of the unknown system parameters such as process and measurement noise. The gating network performs on-line adaptation of the weights given to individual filter estimates ba...

In essence, designing a perfect-reconstruction (PR) biorthogonal cosine-modulated filter bank (BCM) is a non-convex constrained optimization problem that can be solved in principle using general optimization solvers. However, when the number of channels is large and the order of the prototype filter (PF) is high, numerical difficulties in using those optimization solvers often occur, and the co...

Designing appropriate features for acoustic event recognition tasks is an active field of research. Expressive features should both improve the performance of the tasks and also be interpret-able. Currently, heuristically designed features based on the domain knowledge requires tremendous effort in hand-crafting, while features extracted through deep network are difficult for human to interpret...

In many applications such as wireless communications and subband adaptive filtering, we need to design non-uniform filter banks (NUFB), which may lead to better performances and reduced hardware complexity when compared to uniform filter bank. NUFB satisfying linear phase property for all the constituent filters, are desirable in applications such as speech and image processing and in communica...

In essence, designing a perfect-reconstruction (PR) biorthogonal cosine-modulated filter bank (BCM) is a non-convex constrained optimization problem that can be solved in principle using general optimization solvers. However, when the number of channels is large and the order of the prototype filter (PF) is high, numerical difficulties in using those optimization solvers often occur, and the co...

Quantization watermarking is a method of embedding hidden copyright information based on dithered quantization. Such a scheme is only practical for watermarking applications, where the original signal is available to the detector as for a fingerprinting purpose. The aim of this paper is to analyze the quantization watermarking in the wavelet transform or equivalently in filter bank domain. We i...

NON-UNIFORM NEAR-PERFECT-RECONSTRUCTION OVERSAMPLED DFT FILTER BANKS BASED ON ALLPASS-TRANSFORMS Enisa Galijašević and Jörg Kliewer University of Kiel Institute for Circuits and Systems Theory 24143 Kiel, Germany [email protected], [email protected] ABSTRACT In this contribution we discuss an oversampled non-uniform DFT filter bank, which is derived by allpass frequency transformations from it...

Over the last number of years many different libraries of biorthogonal wavelets have been designed. In addition, the lifting scheme has been developed which unifies the design of biorthogonal wavelets using a simple approach. The lifting scheme allows the custom design of wavelets while maintaining the biorthogonal constraint on the filter bank. A drawback of the lifting scheme is that it does ...

In this paper the application of multirate acoustic echo canceller derived from cochlea model and psychoacustically motivated noise and residual echo attenuation system operating on the signal decomposed in cochlear spaced subbands are described. Polyphase realisation of the FIR filter bank with non-uniform frequency resolution achieved by the frequency transformation of filter characteristic u...

Designing filter banks for source coding purposes classically relies on perfect reconstruction filters. However, several studies have shown recently that taking the quantization noise into account in the design could yield noticeable reduction of the reconstruction distortion. In particular, a joint optimization of the synthesis filters and quantizers with respect to the output mean squared err...