Comparative Analysis of Different Loudness Meters Based on Voice Detection and Gating

After decades of deep investigation, the international broadcasting community represented by technical associations and bodies has set precise standards aimed to objectively assess loudness levels of programmes. Although all standards rely on the same algorithm as described in ITU-R BS1770, there are still two possible ways to implement such metering, including Voice Detection and Gating. These two different implementations might, in some cases, provide measurements that significantly differ from each other. Furthermore, whilst the gating feature is uniquely defined in the updated version of BS1770-3, Voice Detection is not currently specified in any standard and its implementation is independently designed by manufacturers. This paper analysis this scenario by comparing the results and robustness provided by three different loudness meters based on Voice Detection. In addition, those values are compared with measurements obtained by using BS1770-3 compliant loudness meters, including tables, comments, and conclusions. 1. LOUDNESS MEASUREMENT IMPLEMENTATIONS 1.1. ITU-R BS1770 and further releases In 2006 ITU released the first release of BS1770, the algorithm designed to assess the Loudness Level and to measure the True Peak Level of audio programmes in broadcasting. The algorithm is based on the computation of mean-square levels of the audio programme weighted according to the R2LB filtering and the gain levels described in the recommendation [1]. The audio signal is measured on all its duration and the final loudness level is produced by integrating all values gathered during the measurement. This implies that no emphasis is put on any specific sound element, regardless their content or technical characteristic (voice, music, sound effect, intensity, pitch, background or foreground). As said, this method was not able to identify the so called “anchor sound” (typically the voice element of the audio mix) and resulted not being fully effective in case the audio programme presents variations in level such as loudness modulation between background and foreground sounds. Consequently, two approaches have been implemented in order to cope with those limitations and they include: Voice Detection and Gating. The latter is designed to compute the Programme Loudness Level by focusing the measurement on foreground sounds only. This result id obtained by discarding all values falling below a predetermined threshold. In 2012 ITU released an updated version of the its recommendation AES Travaglini Voice detection loudness meters and gating AES 134th Convention, Rome, Italy, 2013 May 4–7 Page 2 of 4 where a gating feature was added by including a relative gating at -10LU [2]. 1.2. Voice detection meters It consists of a logic feature that automatically detects the voice content of a programme and that enables the gathering of loudness levels necessary to compute the final overall value only when those elements are detected. Consequently, only the programme parts that present voice elements will contribute in the computation of the overall loudness level representative of the whole programme. This implementation presents the positive capability to focus the measurement on the “anchor sound” that is mainly used by listeners to assess the perceived loudness level of a content, the speech. However, it also has the limitation to not being usable on content with no or very limited voice content. Furthermore, being based on non-standardized implementations and designs, the measurement obtained with voice detection meters might result less robust than others. This paper primarily aims to investigate this aspect. The first manufacturer to implement a voice detection feature in its loudness meter was Dolby Inc. with the release of its Dialogue Intelligence technology in the Dolby LM100 Broadcast Loudness Meter in 2003. Recently, other companies have pursued the same goal and have included this option in their metering software. This research, aside Dolby LM100, includes the analysis of measurements produced by Nugen Audio VisLM-H and Waves Loudness Meter WLM.