Real-time loudness control for IP-based SPTS and MPTS streams

Abstract

'Loudness Control' is the process of analysing and eventually manipulating the audio components in a digital stream with the aim of maintaining the inter- and intraprogramme average loudness as consistent as possible. Excessive loudness jumps spanning over relatively long periods can be attributed to commercials which use loud volumes as a means of advantage to grab the attention of their target audience. The aim of this research is to provide a better Quality of Experience to the audience, while shifting the aim of commercial producers to partake better content and ideas. The system should transparently, and with as little delay as possible, perform loudness control on a number of TV channels simultaneously and in the most reliable way possible to avoid service disruption which would degrade the Quality of Service provided by the broadcastc1 and the Quality of Experience of the user. The main workflow adopted to achieve this target involved the treatment of TV channels data in multicast form, followed by packet parsing, MPEG audio extraction and MPEG audio decoding. Furthermore, the process of obtaining the statistics follows two main standards, namely, the ITU BS.1770 and the EBU Rl 28 standard, follows. The recommendations defined by these two standards provide guidance on the corrections needed to bring the audio loudness back on target. Given that this research are is relatively new, these standards still carry a main drawback in that the correction algorithms suggested up to now are only capable of performing a single correction once a day. The principle endeavour of this work relates to the development of a faster response mechanism to loudness mismatches. Techniques based on probability density estimation, peak detectors, RMS computers, compression techniques and crest factor computers were used to realise a system that can correct loudness errors in real time. The main achievement of this work consists of a real-time system that is capable of ingesting multicast streams, processing their audio components to match the requirements imposed by the standards, and as per the real-time processor defined above, together with re-transmission of the processed MPEG packets over a new multicast stream for effective delivery to the customers, with overall processing delays smaller than 25 ms. The overall system was also scrutinised through subjective tests to ensure that the processor was not introducing audible artefacts or sever loudness jumps. As per the standard, the final aim is to provide much more loudness-consistent audio streams for an improved Quality of Experience by the target audience.