DTS Clear Dialogue: comprehensible TV dialog thanks to machine learning

Xperi's DTS brand has so far made a name for itself primarily through the (home) cinema sound of the same name and the "DTS Play-Fi" multi-room system. Now the manufacturer wants to significantly improve dialog intelligibility in upcoming televisions under the name "DTS Clear Dialogue". Xperi is presenting a pre-production version of the system at this year's IFA, but only behind closed doors.

Two scenes were shown: an excerpt from the movie "The Martian – Save Mark Watney" and a recording of a TV broadcast of the Tour de France. In "The Martian", the challenge for the algorithm was to filter out the actors' dialog in a scene in which there is strong background noise due to a sandstorm. In the sports broadcast, the aim was to isolate the commentators' voices while leaving the cheers of the spectators on the track in the background.

This worked perfectly in both cases. The dialog was much easier to understand – and the volume could be individually adjusted using the remote control. The voices did not sound hollow, as is the case with some previous dialog enhancers. AI algorithms are used for this; Xperi itself speaks of machine learning. Such a result would therefore not be possible using conventional means alone, for example by boosting certain frequencies.

When asked, the presenter also explained that DTS Clear Dialogue does not interfere with the dynamics or change the intention of a scene, for example by raising all dialog to the same volume. A whisper therefore remains a whisper, just as a scream remains a scream. According to the company, it also doesn't matter which language is being spoken or whether the source is a TV broadcast, streaming or Blu-ray disc.

Simulation games

According to Xperi, it is currently in talks with several processor and TV manufacturers to get the technology into televisions as quickly as possible. However, the company can not yet provide specific names or an exact time frame; in the best-case scenario, Xperi believes that the first TV models could be equipped with DTS Clear Dialogue as early as next year.

Xperi is initially targeting televisions. However, if it is successful in this area, it says it is quite conceivable that soundbars and audio/video receivers will also be equipped with the system.

In itself, the problem of poor dialog intelligibility could be a thing of the past. With Dolby AC-4 and MPEG-H Audio, two audio codecs have been available for the broadcast sector for a number of years that allow dialog to be stored and transmitted separately and then regulated separately from the rest on the TV. Back in 2016, c't published the article "New sounds on TV - audio codecs for the television of the future".

(nij)