Bernd Michaelis

A Hidden Markov Model-based continuous gesture recognition system for hand motion trajectory

In this paper, we propose an automatic system that recognizes both isolated and continuous gestures for Arabic numbers (0-9) in real-time based on hidden Markov model (HMM). To handle isolated gestures, HMM using ergodic, left-right (LR) and left-right banded (LRB) topologies with different number of states ranging from 3 to 10 is applied. Orientation dynamic features are obtained from spatio-temporal trajectories and then quantized to generate its codewords. The continuous gestures are recognized by our novel idea of zero-codeword detection with static velocity motion. Therefore, the LRB topology in conjunction with forward algorithm presents the best performance and achieves average rate recognition 98.94% and 95.7% for isolated and continuous gestures, respectively.

Optical 3D Surface Reconstruction by a Multi-Period Phase Shift Method

One problem of classical phase shifting for optical 3D surface reconstruction is the occurrence of ambiguities due to the use of fringe projection. We generally derive a number-theoretical approach to calculate absolute phase measurements which can be used as a base for a reliable surface reconstruction without any ambiguity. The essence of our method is the application of pattern sequences with different periods whereby we homogeneously use all pictures which were taken for the measurement. This leads to a higher average accuracy in the surface reconstruction. Furthermore we propose a technique to avoid typical calculation errors that are produced in classical phase shifting caused by discontinuities, occlusions and reflections on the surface.

Hand trajectory-based gesture spotting and recognition using HMM

In this paper, we propose an automatic system that executes hand gesture spotting and recognition simultaneously without any time delay based on Hidden Markov Models (HMM). Our system is based on three main stages; preprocessing, feature extraction and classification. In preprocessing stage, color and 3D depth map are used to detect hands. The hand trajectory will take place in further steps using Mean-shift algorithm and Kalman filter. The second stage, Orientation dynamic features are obtained from spatio-temporal trajectories and then are quantized to generate its codewords. In the final stage, the gestures are segmented by finding the start and the end points of meaningful gestures that are embedded in the input stream and then are recognized by Viterbi algorithm. Experimental results demonstrate that, our system can successfully recognize spotted hand gestures with a 95.87% recognition rate for Arabic numbers from 0 to 9.

A Novel Method for 3D Face Detection and Normalization

When automatically analyzing images of human faces, either for recognition in biometry applications or fa- cial expression analysis in human machine interaction, one has to cope with challenges caused by different head pose, illumination and expression. In this article we propose a new stereo based method for effectively solving the pose problem through 3D face detection and normalization. The proposed method applies a model-based matching and is especially intended for the study of facial features and the description of their dynamic changes in image sequences under the assumption of non-cooperative persons. In our work, we are currently implementing a new application to observe and analyze single faces of post-operative patients. In the proposed method, face detection is based on color driven clustering of 3D points derived from stereo. A mesh model is matched with the post-processed face cluster using a variant of the Iterative Closest Point algorithm (ICP). Pose is derived from correspondence. Then, pose and model in- formation is used for the synthesis of the face normalization. Results show, stereo and color are powerful cues for finding the face and its pose under a wide range of poses, illumina- tions and expressions (PIE). Head orientation may vary in out of plane rotations up to ±45°. Key words—Image and Video Processing, ICP-Matching, Computer Vision, 3D Face Detection, Normalization

A framework for the integration of gesture and posture recognition using HMM and SVM

For a successful real-time vision-based HCI system, inference from natural visual method is crucial. In this paper, we have aimed to provide interaction through gesture and posture recognition for alphabets and numbers. In addition, data fusion is carried out which integrates these systems to extract multiple meanings at the same time. 3D information is exploited for segmentation and detection of face and hands using normal Gaussian distribution and depth information. For gesture, orientation of two consecutive hand centroid points is computed which is then quantized to generate code words. HMM is trained by Baum Welch algorithm and classified by Viterbi path algorithm. In posture recognition, American Sign Language is recognized for static alphabets and numbers. Feature vectors are computed from statistical and geometrical properties of the hand and are used to train SVM for classification and recognition. Moreover, curvature analysis is carried out for alphabets to avoid misclassifications. Experimental results of the proposed framework successfully integrate both gesture and posture recognition system at decision level fusion whereas the gesture system achieves recognition rate of 98% (i.e. for alphabets and numbers) and the posture recognition system with recognition rates of 98.65% and 98.6% for ASL alphabets and numbers respectively.

Spike-timing-dependent plasticity in small-world networks

Biologically plausible excitatory neural networks develop a persistent synchronized pattern of activity depending on spontaneous activity and synaptic refractoriness (short term depression). By fixed synaptic weights synchronous bursts of oscillatory activity are stable and involve the whole network. In our modeling study we investigate the effect of a dynamic Hebbian-like learning mechanism, spike-timing-dependent plasticity (STDP), on the changes of synaptic weights depending on synchronous activity and network connection strategies (small-world topology). We show that STDP modifies the weights of synaptic connections in such a way that synchronization of neuronal activity is considerably weakened. Networks with a higher proportion of long connections can sustain a higher level of synchronization in spite of STDP influence. The resulting distribution of the synaptic weights in single neurons depends both on the global statistics of firing dynamics and on the number of incoming and outgoing connections.

A Robust Method for Hand Gesture Segmentation and Recognition Using Forward Spotting Scheme in Conditional Random Fields

This paper proposes a forward spotting method that handles hand gesture segmentation and recognition simultaneously without time delay. To spot meaningful gestures of numbers (0-9) accurately, a stochastic method for designing a non-gesture model using Conditional Random Fields (CRFs) is proposed without training data. The non-gesture model provides a confidence measures that are used as an adaptive threshold to find the start and the end point of meaningful gestures. Experimental results show that the proposed method can successfully recognize isolated gestures with 96.51% and meaningful gestures with 90.49% reliability.

Locally Adaptive Fuzzy Image Enhancement

In recent years, some researchers have applied the concept of fuzziness to develop new enhancement algorithms. The global fuzzy image enhancement methods, however, fail occasionally to achieve satisfactory results. In this work, we introduce a locally adaptive version of two existing fuzzy image enhancement algorithms to overcome this problem.

Robust methods for hand gesture spotting and recognition using Hidden Markov Models and Conditional Random Fields

This paper proposes an automatic method that handles hand gesture spotting and recognition simultaneously. To spot meaningful gestures of numbers (0–9) accurately, a stochastic method for designing a non-gesture model with Hidden Markov Models (HMMs) versus Conditional Random Fields (CRFs) is proposed without training data. The non-gesture model provides a confidence measure that is used as an adaptive threshold to find the start and the end point of meaningful gestures, which are embedded in the input video stream. To reduce the states number of the non-gesture model with HMMs, similar probability distributions states are merged based on relative entropy measure. Additionally, the weights of self-transition feature functions are increased for short gesture to further improve the accuracy of gesture spotting and recognition with CRFs. Experimental results show that; the proposed method can successfully spot and recognize meaningful gestures with 93.31% and 90.49% reliability for HMMs and CRFs respectively. In addition, the model inference by HMMs are faster and the saving time is 66.42% using relative entropy. The reliability of CRFs method is improved from 86.12% to 90.49% using short gesture detector.

Data Gathering for Gesture Recognition Systems Based on Mono Color-, Stereo Color- and Thermal Cameras

In this paper, we present our results to build an automatic gesture recognition system using different types of cameras to compare them in reference to their features for segmentation. Normally, the images of a mono color camera system are mostly used as input data in the research area of gesture recognition. In comparison to that, the analysis results of a stereo color camera and a thermal camera system are used to determine the advantages and disadvantages of these camera systems. With this basics, a real-time gesture recognition system is build to classify alphabets (A-Z) and numbers (0-9) with an average recognition rate of 98% using Hidden Markov Models (HMM).