No description available.

No description available.

No description available.

No description available.

No description available.

No description available.

Automatic speech recognition technology is becoming increasingly widespread in many applications. For dictation tasks, where a single talker is to use the system for long periods of time, the high recognition accuracies obtained are in part due to the user performing a lengthy enrolment procedure to ‘tune’ the parameters of the recogniser to their particular voice characteristics and speaking style. Interactive speech systems, where the speaker is using the system for only a short period of time (for example to obtain information) do not have the luxury of long enrolments and have to adapt rapidly to new speakers and speaking styles. This thesis discusses the variations between speakers and speaking styles which result in decreased recognition performance when there is a mismatch between the talker and the systems models. An unsupervised method to rapidly identify and normalise differences in vocal tract length is presented and shown to give improvements in recognition accuracy for little computational overhead. Two unsupervised methods of identifying speakers with similar speaking styles are also presented. The first, a data-driven technique, is shown to accurately classify British and American accented speech, and is also used to improve recognition accuracy by clustering groups of similar talkers. The second uses the phonotactic information available within pronunciation dictionaries to model British and American accented speech. This model is then used to rapidly and accurately classify speakers.

No description available.

This thesis presents an account of an investigation into the use of information theory measures in pattern recognition problems. The objectives were firstly to determine the information content of the set of representations of an input image which are found at the output of an array of sensors; secondly to assess the information which may be used to allocate different patterns to appropriate classes in order to provide a means of recognition; and thirdly to assess the recognition capability of pattern recognition systems and their efficiency of utilization of information. Information assessment techniques were developed using fundamental principles of information theory. These techniques were used to assess the information associated with attributes such as orientation and location, of a variety of input images. The techniques were extended to permit the assessment of recognition capability and to provide a measure of the efficiency with which pattern recognition systems use the information available.

The result of many years of international co-operation in video coding has been the development of algorithms that remove interframe redundancy, such that only changes in the image that occur over a given time are encoded for transmission to the recipient. The primary process used here is the derivation of pixel differences, encoded in a method referred to as Differential Pulse-Coded Modulation (DPCM)and this has provided the basis of contemporary research into low-bit rate hybrid codec schemes. There are, however, instances when the DPCM technique cannot successfully code a segment of the image sequence because motion is a major cause of interframe differences. Motion Compensation (MC) can be used to improve the efficiency of the predictive coding algorithm. This thesis examines current thinking in the area of motion-compensated video compression and contrasts the application of differing algorithms to the general requirements of interframe coding. A novel technique is proposed, where the constituent features in an image are segmented, classified and their motion tracked by a local search algorithm. Although originally intended to complement the DPCM method in a predictive hybrid codec, it will be demonstrated that the evaluation of feature displacement can, in its own right, form the basis of a low bitrate video codec of low complexity. After an extensive discussion of the issues involved, a description of laboratory simulations shows how the postulated technique is applied to standard test sequences. Measurements of image quality and the efficiency of compression are made and compared with a contemporary standard method of low bitrate video coding.