Audio coding and identification for an interactive television application

• Main Author: Sheets, Gregory S.
• Other Authors: Electrical Engineering
• Format: Others
• Language: en
• Published: Virginia Tech 2014
• Subjects: LD5655.V855 1996.S543
• Online Access: http://hdl.handle.net/10919/41071; http://scholar.lib.vt.edu/theses/available/etd-02132009-172049/

This thesis describes the development of the audio coding and identification algorithms for an Interactive Video Data Service (IVDS) system. The overall purpose of the IVDS system that is discussed in this thesis is to provide the user with a wireless return path to the host site. The information being broadcast from the host site can vary. However, it is assumed here that the host site is broadcasting some type of advertisement or commercial, in which case the user's return path is for communicating an interest to the host site in obtaining something of value. The host site is made aware of the interest of the user when a transmission from the user is received. After the user's transmission has been obtained, decoded, and analyzed the host site takes an appropriate course of action. Our particular IVDS system must be noninvasive to the existing broadcast system and have no physical contact with the television at the user end of the system. Our system must also allow the user to make a selection at virtually any time. The performance and feasibility of such a system is studied. The necessary number of bits per user transmission, speed of commercial identification, and the success rate of the commercial identification process are tradeoff factors in our IVDS application. We examine these tradeoff factors and discuss the effect various approaches to solving our problem have on them. The correlation coefficient is used as a distance measure in our IVDS application and compared on the basis of feasibility to other distance measures that have recently been used in the related problems of speech recognition and speaker identity verification. The performance of a few distance measures that are based on the inner product are studied through various simulations. The performance of the ‘optimal’ system that is developed as a result of the various simulations is acceptable for 'small scale' usage. The performance can be optimized for various criteria and depends on the distance measure used and some of the parameters of the proposed system. === Master of Science