Machine learning algorithms have shown their abilities to tackle difficult recognition problems, but they are still rife with challenges. Among these challenges is how to deal with problems where new categories constantly occur, and the datasets can dynamically grow. Most contemporary learning algorithms developed to this point are governed by the assumptions that all testing data classes must be the same as training data classes, often with equal distribution. Under these assumptions, machine-learning algorithms can perform very well, using their ability to handle large feature spaces and classify outliers. The systems under these assumptions are called Closed Set Recognition systems (CSR). However, these assumptions cannot reflect practical applications in which out-of-set data may be encountered. This adversely affects the recognition prediction performances. When samples from a new class occur, they will be classified as one of the known classes. Even if this sample is far from any of the training samples, the algorithm may classify it with a high probability, that is, the algorithm will not only be wrong, but it may also be very confident in its results. A more practical problem is Open Set Recognition (OSR), where samples of classes not seen during training may show up at testing time. Inherently, there is a problem how the system can identify the novel sound classes and how the system can update its models with new classes. This thesis highlights the problems of multi-class recognition for OSR of sounds as well as incremental model adaptation and proposes solutions towards addressing these problems. The proposed solutions are validated through extensive experiments and are shown to provide improved performance over a wide range of openness values for sound classification scenarios.

Finger gesture recognition using glove-like interfaces are very accurate for sensing individual finger positions by employing a gamut of sensors. However, for the same reason, they are also very costly, cumbersome and unaesthetic for use in artistic scenarios such as gesture based music composition platforms like Virginia Tech's Linux Laptop Orchestra. Wearable computing has shown promising results in increasing portability as well as enhancing proprioceptive perception of the wearers' body. In this thesis, we present the proof-of-concept for designing a novel muscle-machine interface for interpreting human thumb motion as a 2-dimensional joystick employing mechanomyographic signals. Infrared camera based systems such as Microsoft Digits and ultrasound sensor based systems such as Chirp Microsystems' Chirp gesture recognizers are elegant solutions, but have line-of-sight sensing limitations. Here, we present a low-cost and wearable joystick designed as a wristband which captures muscle sounds, also called mechanomyographic signals. The interface learns from user's thumb gestures and finally interprets these motions as one of the four kinds of thumb movements. We obtained an overall classification accuracy of 81.5% for all motions and 90.5% on a modified metric. Results obtained from the user study indicate that mechanomyography based wearable thumb-joystick is a feasible design idea worthy of further study. / Master of Science

No description available.

Digitized by Kansas Correctional Industries

This research focuses on iris recognition, the most accurate form of biometric identification. The robustness of iris recognition comes from the unique characteristics of the human, and the permanency of the iris texture as it is stable over human life, and the environmental effects cannot easily alter its shape. In most iris recognition systems, ideal image acquisition conditions are assumed. These conditions include a near infrared (NIR) light source to reveal the clear iris texture as well as look and stare constraints and close distance from the capturing device. However, the recognition accuracy of the-state-of-the-art systems decreases significantly when these constraints are relaxed. Recent advances have proposed different methods to process iris images captured in unconstrained environments. While these methods improve the accuracy of the original iris recognition system, they still have segmentation and feature selection problems, which results in high FRR (False Rejection Rate) and FAR (False Acceptance Rate) or in recognition failure. In the first part of this thesis, a novel segmentation algorithm for detecting the limbus and pupillary boundaries of human iris images with a quality assessment process is proposed. The algorithm first searches over the HSV colour space to detect the local maxima sclera region as it is the most easily distinguishable part of the human eye. The parameters from this stage are then used for eye area detection, upper/lower eyelid isolation and for rotation angle correction. The second step is the iris image quality assessment process, as the iris images captured under unconstrained conditions have heterogeneous characteristics. In addition, the probability of getting a mis-segmented sclera portion around the outer ring of the iris is very high, especially in the presence of reflection caused by a visible wavelength light source. Therefore, quality assessment procedures are applied for the classification of images from the first step into seven different categories based on the average of their RGB colour intensity. An appropriate filter is applied based on the detected quality. In the third step, a binarization process is applied to the detected eye portion from the first step for detecting the iris outer ring based on a threshold value defined on the basis of image quality from the second step. Finally, for the pupil area segmentation, the method searches over the HSV colour space for local minima pixels, as the pupil contains the darkest pixels in the human eye. In the second part, a novel discriminating feature extraction and selection based on the Curvelet transform are introduced. Most of the state-of-the-art iris recognition systems use the textural features extracted from the iris images. While these fine tiny features are very robust when extracted from high resolution clear images captured at very close distances, they show major weaknesses when extracted from degraded images captured over long distances. The use of the Curvelet transform to extract 2D geometrical features (curves and edges) from the degraded iris images addresses the weakness of 1D texture features extracted by the classical methods based on textural analysis wavelet transform. Our experiments show significant improvements in the segmentation and recognition accuracy when compared to the-state-of-the-art results.

This thesis introduces a general method for using information at the utterance level and across utterances for automatic speech recognition. The method involves classification of utterances into types. Using constraints at the utterance level via this classification method allows information sources to be exploited which cannot necessarily be used directly for word recognition. The classification power of three sources of information is investigated: the language model in the speech recogniser, dialogue context and intonation. The method is applied to a challenging task: the recognition of spontaneous dialogue speech. The results show success in automatic utterance type classification, and subsequent word error rate reduction over a baseline system, when all three information sources are probabilistically combined.

No description available.

No description available.

No description available.

No description available.