Tobias Dahl

A bridge between Tucker-1 and Carroll's generalized canonical analysis

A set of tools for analyzing relationships between and within multiple data matrices is presented. The starting point is a new, unified approach bridging two existing methods; Carrolls Generalized Canonical Analysis (GCA) with the Tucker-1 method for principal component analysis of multiple matrices. GCA and Tucker-1 are shown to correspond to particular choices of a ridge parameter. The unified method is then generalized to a larger space of methods.

Outlier and group detection in sensory panels using hierarchical cluster analysis with the Procrustes distance

Generalised Procrustes analysis (GPA) is a much-used method for analysing sensory profile data. In this paper, hierarchical clustering using the Procrustes distance is proposed for situations where the data profiles are believed to come from a non-homogeneous group. This new approach to sensory panel analysis may be used at an exploratory stage, in combination with GPA, to gain insight into the structures of the data set. It can help the researcher detect outliers and subgroups, help him/her make decisions regarding further analysis, and reduce the risk of erroneous inference about the data.

BIMA: Blind iterative MIMO algorithm

Identification of the channel matrix is of main concern in wireless MIMO (Multiple Input Multiple Output) systems. Here, we present an SVO-based approach for blind identification of the main independent parallel channels. The right and left singular vectors are estimated directly (no channel matrix estimation is necessary) and continuously updated during normal transmission. The approach is related to the iterative Power Method (8), as well as the time reversal approach ([4]).

Applications of airborne ultrasound in human-computer interaction.

Airborne ultrasound is a rapidly developing subfield within human-computer interaction (HCI). Touchless ultrasonic interfaces and pen tracking systems are part of recent trends in HCI and are gaining industry momentum. This paper aims to provide the background and overview necessary to understand the capabilities of ultrasound and its potential future in human-computer interaction. The latest developments on the ultrasound transducer side are presented, focusing on capacitive micro-machined ultrasonic transducers, or CMUTs. Their introduction is an important step toward providing real, low-cost multi-sensor array and beam-forming options. We also provide a unified mathematical framework for understanding and analyzing algorithms used for ultrasound detection and tracking for some of the most relevant applications.

Intrinsic Subspace Convergence in TDD MIMO Communication

In numerical linear algebra, students encounter early the iterative power method, which finds eigenvectors of a matrix from an arbitrary starting point through repeated normalization and multiplications by the matrix itself. In practice, more sophisticated methods are used nowadays, threatening to make the power method a historical and pedagogic footnote. However, in the context of communication over a time-division duplex (TDD) multiple-input multiple-output (MIMO) channel, the power method takes a special position. It can be viewed as an intrinsic part of the uplink and downlink communication switching, enabling estimation of the eigenmodes of the channel without extra overhead. Generalizing the method to vector subspaces, communication in the subspaces with the best receive and transmit signal-to-noise ratio (SNR) is made possible. In exploring this intrinsic subspace convergence (ISC), we show that several published and new schemes can be cast into a common framework where all members benefit from the ISC

Design of acoustic lenses for ultrasonic human-computer interaction

A novel Twisted Acoustic Lens (TAL) design to spread the radiated acoustic energy of a flat transducer is proposed. The lens is comprised of many perforations deployed uniformly, with trajectories that are circular helices of different radii. We have fabricated and characterized a disc-shaped TAL of thickness 7 mm and twist angle of 45° and demonstrated that the energy radiated by a directive ultrasonic transducer can be significantly dispersed. It is possible to almost triple the -6 dB directivity of a circular transducer of 20 mm of diameter at frequencies between 60 kHz and 80 kHz. The proposed lens design has great potential in certain ultrasonic touchless applications in which a commercially available ultrasonic transducer needs to be turned into field-of-view maximizing acoustic component while maintaining a flat front end.

High-resolution ultrasonic gesture tracking systems for future portable devices

In the present paper we explore transducer housing designs for high-resolution ultrasonic gesture tracking systems (UGTS) in future portable devices. We show characterization results of two design concepts based on the use of diffractive properties of ultrasound in order to maximize the sensing volume useful for touchless interaction. The designs are described in detail, a validation prototype was constructed and evaluated both using simulations and measurements. In doing so, we develop some simple yet relevant numerica models for the diffractive slits. The proposed design ideas have the potential to be implemented in future UGTS while maintaining sufficiently low friction in the industrial design process, sufficiently low cost and high performance in the user experience.

Local-ICA: independent component analysis for sources with non-stationary amplitudes

We propose a novel and conceptually simple method for blind source separation of linear mixtures. Sources are required to have independently fluctuating non-stationary amplitudes, where the latter are defined relative to a moving time frame. Utilizing the local temporal structure of the mixtures, the proposed method estimates the mixing matrix and the sources by repeated use of the singular value decomposition and k-means clustering. Performance is comparable to FastICA [2] on some typical test problems. Unlike the ICA algorithms that are based on higher order statistics, LocalICA is based only on second order statistics. LocalICA thus has the potential for separating mixtures of colored Gaussian sources.