Digital Audio Broadcasting Proposal

Introduction

Real-time multimedia services contributed largely to the life and according to numerous estimations, showed great expansion through the Internet domain. However, the diversity of information and communication technology (ICT) was often narrowed to computers and the Internet [1]. New ICTs were considered as autonomous from previous ones while strategies and programs that combined them held more promise. For instance, the older medium radio was given less attention than during the past decades. Its history was long, and its presence in many people’s lives was quite normal, yet was often overlooked and ignored. Radio, however, could still offer many interesting solutions of its appliance in that modern age of digital technologies [2]. I believe that assimilation of modern ICTs with older communication technologies (e.g. radio) could be more sufficient than without. During the course of that proposal, I tried to persuade the reader that the potential of radio to offer cheap and effective communication channels in sparsely populated or less developed regions was underestimated. Radio could not lose its role in audio world, at where you could have your own privacy [3], at where you could listen with simple devices to useful and entertaining information - everywhere with no extra costs. Thus, I tried to create my own digital radio station.

However, the development of digital radio was very variable with some nations having a large number of digital radio services across the whole country while others had very few services and devices [4]. Therefore, in order to succeed with that endeavor of mine, I had to provide my online radio station with listeners. How did I do it?

Interface

Whittaker et al. evaluated interfaces to support navigation within speech documents providing a mixed solution for global and local navigation by a graphical interface in combination with an audio output [5]. Emnett and Schmandt developed a system that searches for story boundaries in news broadcasts and enabled a nonlinear navigation in the audible content with the help of different interface approaches [6].

The problems of mobile devices in rough environments in relation with a speech-based interaction were reported in [7]. A notification system “dynamically selected the relevant presentation level for incoming messages” (such as email, voice mails, news broadcasts) based on priority and user activity. Users could browse these messages by means of tactile and speech recognition input on a wearable audio computing platform. Recent developments for Large Vocabulary Continuous Speech Recognition (LVCSR) systems used methods of neural networks and of cepstral coefficients [8].

A comprehensive overview of current research topics on advanced Speech Dialogue Systems (SDS) was given in [9]. The feasibility in mobile devices with the help of distributed approaches was shown in [10]. There were prototypes for web-based SD85 and audio search engines that were able to search spoken words of podcasts and radio signals for queries entered with a keyboard [11] [12]. The Voice XML standard could be used to abstract the dialog constructs, the user input/output, the control flow, the environment, and resources [13].

Fundamentals

In this section, I introduced the information sources and approaches, which were useful in improvement and establishment of the system. I analyzed the various data transmission techniques, presented a survey of the available information sources in the Digital Radio environment, and introduced methods related to the Information Retrieval (IR) process such as the Automatic Speech Recognition (ASR), Music-Speech Discrimination (MSD) respectively speech extraction, and Text Processing (TP) as well as Text-to-Speech (TTS) procedures for the output of spoken information.

Information Sources

Two general types of information sources could be distinguished for DAB:

1) The data services (service information. additional data services) were available as text, Internet-based information was applicable, as well. Primary sources of information for the receiver were service- and program-related data, comparable with FM-radio data system (RDS). Broadcast Websites (BWS) contained multifaceted news, press reviews etc. Other sources of information were Dynamic Label Plus (DL), and Electronic Program Guides (EPG), where providers were able to transmit supplementary, as well as program-independent information.

2) The audible signals were after MSD converted by an ASR into plain text in order to perform a content-based analysis. Internet audio and podcast files were applicable, as well. Recorded information could be of any kind with regards to content e.g. breaking news, headlines, educational and cultural items, current affairs, discussions, etc.

Metadata for Spoken Content

The functionality of IR systems was highly dependent on the usage of appropriate Metadata. It was a difficult task, especially for spoken content, to come to a compromise between accuracy of description, universal adaptability, and computational overhead. It was possible to apply subsets of the TV-Anytime standard [16] offering XML-structures feasible in the field of search, recommender and archive applications [17]. There were a couple of different standards and approaches for DAB defining how to transmit additional information beside the actual audio program. An early proposal for a “technique of machine-interpretable DAB” contained description and receiver hardware control, concerning the appliance and using of Dynamic Label (DL) fields in the transmitted frames, was formulated by Nathan et al. in [18]. A similar approach was chosen in [19]. The separation of information was carried out in both cases by machine-readable control characters.

The complex EPG used TV-Anytime and MPEG-7 metadata with an XML-based data structure. The internal structure allowed describing content elements in a hierarchical structure with any depth and level of granularity according to numerous descriptors. The defined catchwords and descriptors were derived from a controlled vocabulary that is capacious, but of fixed extent. The TV-Anytime standard embraced a comprehensive collection of classification schemes. Those schemes consisted of elaborate arrangements containing pre-assigned terms that were used as catchwords to attach several categories to AV-contents.