Contents I. Introduction
The digital multi-channel TV broadcasting services in Japan, which started in December 2000, provide a wide range of information services such as a huge number of TV program channels with digital and analog forms, electronic program tables' distribution service and data broadcasting services (e.g., weather forecasts). However, while the TV has become multifunctional, its operation has become complex. With the further increase of the range as well as the variety of information services, the TV set will increase its multifunctionality and its operation will become even more complex. Therefore we proposed a speech recognition system as a digital TV (DTV) interface and released a commercial product
Panasonic BS-Digital Hi-Vision TV, TH-36DH200 was a product designed for the Japanese market and released on Dec. 2001
. Considering a speech recognition system, for home application, the automatic speech recognition (ASR) systems must be simple to use and robust to speaker variations caused by generation diversity such as children and aged persons. To provide speaker variation robustness, we developed a speaker adaptation technique, a normalization technique with frequency-warping procedure employed with an age-dependent acoustic model. With the implementation of this technique and a design of an access-friendly interface, “DTV with Speech Recognition Remote Control” was developed. In the following sections, we introduce its user interface, the speech recognition system, the speaker adaptation technique and their performance.