Showing papers by "Hazim Kemal Ekenel published in 2005"

Local appearance based face recognition using discrete cosine transform

Abstract: In this paper, a local appearance based face recognition algorithm is proposed. In the proposed algorithm local information is extracted using block-based discrete cosine transform. Obtained local features are combined both at the feature level and at the decision level. The performance of the proposed algorithm is tested on the Yale and CMU PIE face databases, and the obtained results show significant improvement over the holistic approaches.

Kalman filters for audio-video source localization

Abstract: In prior work, we proposed using an extended Kalman filter to directly update position estimates in a speaker localization system based on time delays of arrival. We found that such a scheme provided superior tracking quality as compared with the conventional closed-form approximation methods. In this work, we enhance our audio localizer with video information. We propose an algorithm to incorporate detected face positions in different camera views into the Kalman filter without doing any explicit triangulation. This approach yields a robust source localizer that functions reliably both for segments wherein the speaker is silent, which would be detrimental for an audio only tracker, and wherein many faces appear, which would confuse a video only tracker. We tested our algorithm on a data set consisting of seminars held by actual speakers. Our experiments revealed that the audio-video localizer functioned better than a localizer based solely on audio or solely on video features.

Feature weighted mahalanobis distance: Improved robustness for Gaussian classifiers

Abstract: Gaussian classifiers are strongly dependent on their underlying distance method, namely the Mahalanobis distance. Even though widely used, in the presence of noise this distance measure loses dramatically in performance, due to equal summation of the squared distances over all features. The features with large distance can mask all the other features so that the classification considers only these features, neglecting the information provided by the other features. To overcome this drawback we propose to weight the different features in the Mahalanobis distance according to their distances after the variance normalization. The idea behind this is to give less weight to noisy features and high weight to noise free features which are more reliable. Thereafter, we replace the traditional distance measure in a Gaussian classifier with the proposed. In a series of experiments we show the improved noise robustness of Gaussian classifiers by the proposed modifications in contrast to the traditional approach.

The connector: facilitating context-aware communication

Abstract: We present the Connector, a context-aware service that intelligently connects people. It maintains an awareness of its users' activities, preoccupations and social relationships to mediate a proper connection at the right time between them. In addition to providing users with important contextual cues about the availability of potential callees, the Connector adapts the behavior of the contactee's device automatically in order to avoid inappropriate interruptions.To acquire relevant context information, perceptual components analyze sensor input obtained from a smart mobile phone and --- if available --- from a variety of audio-visual sensors built into a smart meeting room environment. The Connector also uses any available multimodal interface (e.g. a speech interface to the smart phone, steerable camera-projector, targeted loudspeakers) in the smart meeting room, to deliver information to users in the most unobtrusive way possible.

Multi-modal person recognition for vehicular applications

Abstract: In this paper, we present biometric person recognition experiments in a real-world car environment using speech, face, and driving signals. We have performed experiments on a subset of the in-car corpus collected at the Nagoya University, Japan. We have used Mel-frequency cepstral coefficients (MFCC) for speaker recognition. For face recognition, we have reduced the feature dimension of each face image through principal component analysis (PCA). As for modeling the driving behavior, we have employed features based on the pressure readings of acceleration and brake pedals and their time-derivatives. For each modality, we use a Gaussian mixture model (GMM) to model each person's biometric data for classification. GMM is the most appropriate tool for audio and driving signals. For face, even though a nearest-neighbor-classifier is the preferred choice, we have experimented with a single mixture GMM as well. We use background models for each modality and also normalize each modality score using an appropriate sigmoid function. At the end, all modality scores are combined using a weighted sum rule. The weights are optimized using held-out data. Depending on the ultimate application, we consider three different recognition scenarios: verification, closed-set identification, and open-set identification. We show that each modality has a positive effect on improving the recognition performance.

A generic face representation approach for local appearance based face verification

Abstract: In this paper we present the experimental results of a generic local appearance based face representation approach obtained from the first and fourth experiments of the Face Recognition Grand Challenge (FRGC) version 1 data. The introduced representation approach is compared with the baseline system with the standard distance metrics of L1 norm, L2 norm and cosine angle. The experimental results show that the proposed local appearance based approach provides better and more stable results than the baseline system -holistic Eigenfaces- approach.

Multimodal person verification from video sequences

Abstract: In this paper, a multimodal person verification system based on fusing information derived from face speech signals is proposed. Principle component analysis and independent component analysis techniques are used for face verification and melfrequency-cepstral coefficients are used for speaker verification. The matching scores from individual modalities are combined using the sum rule. The results indicate that fusing indivual modalities improve overall performance of the verification system.