Edwige Pissaloux

A compact tactile display for the blind with shape memory alloys

This paper presents a novel tactile display device that exhibits some advantageous features for the blind: low-cost, lightweight, compactness and high portability. The prototype consists of an 8 times 8 arrangement of tactile pins actuated by shape memory alloys (SMAs) with 2.6 mm spatial resolution and 1 mm vertical excursion. Unlike existing devices, the full display is only 200 g weight and its compact dimensions make it easily carried by the user. A technical overview of the system is presented as well as preliminary results on tactile perception to evaluate its performance on information transmission

Gaze estimation using local features and non-linear regression

In this paper, we present an appearance-based gaze estimation method for a head-mounted eye tracker. The idea is to extract discriminative image descriptors with respect to gaze before applying a regression scheme. We employ multilevel Histograms of Oriented Gradients (HOG) features as our appearance descriptor. To learn the mapping between eye appearance and gaze coordinates, two learning-based approaches are evaluated : Support Vector Regression (SVR) and Relevance Vector Regression (RVR). Experimental results demonstrate that, despite the high dimensionality, our method works well and RVR provides a more efficient and generalized solution than SVR by retaining a low number of basis functions.

Causality inversion in the reproduction of roughness

When a finger scans a non-smooth surface, a sensation of roughness is experienced. A similar sensation is felt when a finger is in contact with a mobile surface vibrating in the tangential direction. Since an actual finger-surface interaction results in a varying friction force, how can a measured friction force can be converted into skin relative displacement. With a bidirectional apparatus that can measure this force and transform it into displacement with unambiguous causality, such mapping could be experimentally established. A pilot study showed that a subjectively equivalent sensation of roughness can be achieved betweem a fixed real surface and a vibrated mobile surface.

Towards activity recognition from eye-movements using contextual temporal learning

Patterns of gaze play an important role in video analysis and understanding human behavior. This paper proposes a novel approach to solve the problem of activity recognition from first-person gaze movements. Novel n-gram statistical features are extracted from these movements via a multi scale and temporal representation scheme. A joint classification and segmentation of activities and a spatio-temporal contextual learning approach built upon confidence values from long-range neighborhoods reinforces the novelty and efficiency of the proposed approach. Experimental results show that the proposed method improves by 18% on the current baseline.

Orientation filtering : a Novel Matching Method

This paper presents a statistical method to match feature points from stereo pairs of images. The proposed method is evaluated in terms of effectiveness, robustness and computational speed. The evaluation was performed on several pairs of real stereo images of natural scenes taken onboard an unmanned aerial vehicle. The results show that the proposed method reduces the number of incorrect matches and is fast.