Rudolf Rabenstein

Nonlinear acoustic echo cancellation with 2nd order adaptive Volterra filters

Acoustic echo cancellers in todays speakerphones or video conferencing systems rely on the assumption of a linear echo path. Low-cost audio equipment or constraints of portable communication systems cause nonlinear distortions, which limit the echo return loss enhancement achievable by linear adaptation schemes. These distortions are a super-position of different effects, which can be modelled either as memoryless nonlinearities or as nonlinear systems with memory. Proper adaptation schemes for both cases of nonlinearities are discussed. An echo canceller for nonlinear systems with memory based on an adaptive second order Volterra filter is presented. Its performance is demonstrated by measurements with small loudspeakers. The results show an improvement in the echo return loss enhancement of 7 dB over a conventional linear adaptive filter. The additional computational requirement for the presented Volterra filter is comparable to that of existing acoustic echo cancellers.

Joint audio-video object localization and tracking

There has been a tremendous amount of research on object localization either involving microphone arrays or video cameras. Considerable less attention has been paid, however, to object localization and tracking based on joint audio-video processing thus far. This may be related to the lack of suitable algorithms for object localization simultaneously using multimicrophone outputs and color image sequences. In this article, we propose a solution to this problem. Before elaborating on joint audio-video processing, we review some previous work the areas of audio and video object localization. Then a recursive sensor fusion method based on decentralized Kalman filtering is introduced. Unfortunately, the decentralized Kalman filter cannot be directly used for joint audio-video object localization due to specific properties of the audio sensor. By properly adjusting the local audio position estimator, however, we manage to keep the overall architecture. We stress the general methodology.

Active listening room compensation for massive multichannel sound reproduction systems using wave-domain adaptive filtering

The acoustic theory for multichannel sound reproduction systems usually assumes free-field conditions for the listening environment. However, their performance in real-world listening environments may be impaired by reflections at the walls. This impairment can be reduced by suitable compensation measures. For systems with many channels, active compensation is an option, since the compensating waves can be created by the reproduction loudspeakers. Due to the time-varying nature of room acoustics, the compensation signals have to be determined by an adaptive system. The problems associated with the successful operation of multichannel adaptive systems are addressed in this contribution. First, a method for decoupling the adaptation problem is introduced. It is based on a generalized singular value decomposition and is called eigenspace adaptive filtering. Unfortunately, it cannot be implemented in its pure form, since the continuous adaptation of the generalized singular value decomposition matrices to the variable room acoustics is numerically very demanding. However, a combination of this mathematical technique with the physical description of wave propagation yields a realizable multichannel adaptation method with good decoupling properties. It is called wave domain adaptive filtering and is discussed here in the context of wave field synthesis.

A real-time face tracker for color video

This paper presents a face localization and tracking algorithm which is based upon skin color detection and principal component analysis (PCA) based eye localization. Skin color segmentation is performed using statistical models for human skin color. The skin color segmentation task results in a mask marking the skin color regions in the actual frame, which is further used to compute the position and size of the dominant facial region utilizing a robust statistics-based localization method. To improve the results of skin color segmentation, a foreground/background segmentation and an adaptive background update scheme are added. Additionally, the derived face position is tracked with a Kalman filter. To overcome the problem of skin color ambiguity, an eye detection algorithm based on PCA is presented.

Steady-state performance limitations of subband adaptive filters

Nonperfect filterbanks used for subband adaptive filtering (SAF) are known to impose limitations on the steady-state performance of such systems. In this paper, we quantify the minimum mean-square error (MMSE) and the accuracy with which the overall SAF system can model an unknown system that it is set to identify. First, in case of MMSE limits, the error is evaluated based on a power spectral density description of aliased signal components, which is accessible via a source model for the subband signals that we derive. Approximations of the MMSE can be embedded in a signal-to-alias ratio (SAR), which is a factor by which the error power can be reduced by adaptive filtering. With simplifications, SAR only depends on the filterbanks. Second, in case of modeling, we link the accuracy of the SAF system to the filterbank mismatch in perfect reconstruction. When using modulated filterbanks, both error limits-MMSE and inaccuracy-can be linked to the prototype. We explicitly derive this for generalized DFT modulated filterbanks and demonstrate the validity of the analytical error limits and their approximations for a number of examples, whereby the analytically predicted limits of error quantities compare favorably with simulations.

Digital sound synthesis by physical modeling using the functional transformation method

1. Introduction.- 2. Sound-Based Synthesis Methods.- 1 Wavetable synthesis.- 1.1 Looping.- 1.2 Pitch shifting.- 1.3 Enveloping.- 1.4 Filtering.- 2 Granular synthesis.- 2.1 Asynchronous granular synthesis.- 2.2 Pitch-synchronous granular synthesis.- 3 Additive synthesis.- 4 Subtractive synthesis.- 5 FM synthesis.- 6 Combinations of sound-based synthesis methods.- 3. Physical Description of Musical instruments.- 1 General notation.- 2 Subdivision of a musical instrument into vibration generators and a resonant body.- 2.1 Division of stringed instruments into single strings and the resonant body.- 2.1.1 Construction of stringed instruments.- 2.1.2 Fixed strings filtered with the resonant body.- 2.1.3 Strings terminated with independent impedances.- 2.1.4 Strings terminated with an impedance network.- 2.2 Division of a kettle drum into a membrane and the kettle.- 2.2.1 Construction of drums.- 2.2.2 Drum body simulation by modifying the physical parameters of the membrane.- 2.2.3 Drum body simulation by room acoustic simulation with the membrane as vibrating boundary.- 3 Physical description of string vibrations.- 3.1 Longitudinal string vibrations.- 3.2 Torsional string vibrations.- 3.3 Transversal string vibrations.- 3.3.1 Basic linear model.- 3.3.2 Nonlinear excitation functions.- 3.3.3 Nonlinear PDE with solution-dependent coefficients.- 4 Physical description of membrane vibrations.- 4.1 Bending membrane vibrations.- 5 Physical description of resonant bodies.- 6 Chapter summary.- 4. Classical Synthesis Methods Based on Physical Models.- 1 Finite difference method.- 1.1 FDM applied to scalar PDEs.- 1.2 FDM applied to vector PDEs.- 2 Digital waveguide method.- 2.1 Digital waveguides simulating string vibrations.- 2.2 Digital waveguide meshes simulating membrane vibrations.- 3 Modal synthesis.- 4 Chapter summary.- 5. Functional Transformation Method.- 1 Fundamental principles of the FTM.- 1.1 FTM applied to scalar PDEs.- 1.1.1 Laplace transformation.- 1.1.2 Sturm-Liouville transformation.- 1.1.3 Transfer function model.- 1.1.4 Discretization of the MD TFM.- 1.1.5 Inverse Sturm-Liouville transformation.- 1.1.6 Inverse z-transformation.- 1.2 FTM applied to vector PDEs.- 1.2.1 Laplace transformation.- 1.2.2 Sturm-Liouville transformation.- 1.2.3 Transfer function model.- 1.2.4 Discretization of the MD TFM.- 1.2.5 Inverse Sturm-Liouville transformation.- 1.2.6 Inverse z-transformation.- 1.3 FTM applied to PDEs with nonlinear excitation functions.- 1.4 FTM applied to PDEs with solution-dependent coefficients.- 1.5 Stability and simulation accuracy of the FTM.- 1.6 Section summary.- 2 Application of the FTM to vibrating strings.- 2.1 Transversal string vibrations described by a scalar PDE.- 2.2 Longitudinal string vibrations described by vector PDEs.- 2.2.1 Boundary conditions of second kind.- 2.2.2 Boundary conditions of third kind.- 2.2.3 Two interconnected strings.- 2.3 Transversal string vibrations with nonlinear excitation functions.- 2.3.1 Piano hammer excitation.- 2.3.2 Slapped bass.- 2.4 Transversal string vibrations with tension-modulated nonlinearities.- 3 Application of the FTM to vibrating membranes.- 3.1 Rectangular reverberation plate.- 3.2 Circular drum heads.- 4 Application of the FTM to resonant bodies.- 5 Chapter summary.- 6. Comparison of the Ftm with the Classical Physical Modeling Methods.- 1 Comparison of the FTM with the FDM.- 2 Comparison and combination of the FTM with the DWG.- 2.1 Comparison of the FTM with the DWG.- 2.2 Combination of the DWG with the FTM.- 2.2.1 Designing the loss filter.- 2.2.2 Designing the dispersion filter.- 2.2.3 Designing the fractional delay filter.- 2.2.4 Adjusting the excitation function.- 2.3 Limits of the combination.- 3 Comparison of the FTM with the MS.- 4 Chapter conclusions.- 7. Summary, Conclusions, and Outlook.

Digital sound synthesis of string instruments with the functional transformation method

The theory of multidimensional continuous and discrete systems is applied to derive a parametric description of musical sounds from a physical model of real or virtual string instruments. The mathematical representation of this model is given by a partial differential equation for a vibrating string. Suitable functional transformations with respect to time and space turn this partial differential equation into a multidimensional transfer function. It is the starting point for the derivation of a discrete-time system by classical analog to discrete transformations. The coefficients of this discrete model depend explicitly on the geometric properties and material constants of the underlying physical model. This ensures a meaningful behaviour of the discrete system under varying conditions and allows for an intuitive control by the user. Furthermore, the performance of real-time implementations is discussed. Finally, several extensions of this synthesis method for computer music applications are presented.

Classification of time delay estimates for robust speaker localization

This paper proposes a solution to the problem of robust speaker localization under adverse acoustic conditions. The approach is based on the classification of time delay estimates. Two classification techniques are investigated in detail: maximum likelihood (ML) classification and classification based on histogram comparison. Their performance under adverse acoustic conditions is compared to outcomes obtained with the traditional approach which uses time delay estimates directly to infer speaker positions. Experiments indicate that the ML classification method provides little improvement over the traditional method. On the other hand, using histogram classification, we can improve the probability of correct speaker localization by more than 60% compared to either the traditional approach or the ML classification technique.

Blocked-based physical modeling for digital sound synthesis

Block-based physical modeling is a methodology for modeling physical systems with different subsystems. It is an important concept for the physical modeling of real or virtual musical instruments where different components may be modeled according to different paradigms. Connecting systems of diverse nature in the discrete-time domain requires a common interconnection strategy. This contribution presents suitable interconnection strategies that incorporate a wide range of modeling blocks and considers the automatic implementation of block-based structures. Software environments are presented, which allow to build complex sound synthesis systems without burdening the user with problems of block compatibility

Interconnection of state space structures and wave digital filters

State space structures (SSSs) and wave digital filters (WDFs) are two major paradigms for the realization of digital filters. Both approaches are well established, but there are no proven methods for a mixed design of digital filters consisting of parts which are realized as SSSs and parts realized as WDFs. This contribution shows how to add a wave port to the conventional SS representation. This wave port allows to interconnect SSSs and WDFs without creating delay-free loops. Such interconnections allow to build discrete-time structures by reusing existing designs from both classes of digital filters.