"Target Microphone" Technology for Sports Broadcasting-- Clear extraction of sports sounds from spectator cheering tested in cooperation with NHK

NTT (Chiyoda-ku, Tokyo, Hiroo Unoura, CEO) has developed 'target microphone technology' for extracting clear sounds of a sporting event, such as the sounds of a strike at the goal in a soccer match, from the mixed sound of spectator cheering.

Since July 2014, NTT has been collaborating with the Japan Broadcasting Corporation (NHK) in testing the target microphone technology on the playing fields of actual sports events. Specifically, we verified that a dynamic sound output that enhances the sense of presence can be achieved by strengthening the sounds of soccer goal shots, the sounds of sumo wrestlers clashing, the sounds of golf shots and the ball dropping, and the sound of a baseball being hit by from 10 to 30 dB (a factor of from 10 to 1000).

In current sports broadcasting, it is possible to zoom a camera in to obtain high-definition video of a desired location. For audio, on the other hand, sound can be picked up with a shotgun microphone or other such means, but the target sound is mixed in with the cheering of spectators, so it is difficult for the viewer to hear the sounds from the playing field that create a sense of presence. The target microphone technology developed by NTT overcomes that problem by delivering the dynamic sounds of competition to the viewer in a way that has never been possible before.

Future plans

NTT plans practical application within one year via the NTT Group companies, and will further investigate audio production for sports broadcasting that provides greater sense of presence by the year of 2020 in collaboration with various partner companies.

NTT will continue to improve this technology to achieve even higher performance in extracting the target sound and to improve sound quality. We will also continue to work with NHK on testing, including other types of sports events, to verify the technical capabilities.

Use scenario

Target Microphone Technology

1.Overview

In sports broadcasting, shotgun microphones are used to pick up distant sounds, but adequate isolation of the target sounds from background noise (cheering by the audience) is not possible with that method.

The target microphone technology can overcome that problem by clearly extracting the sounds of the competition from the signals of multiple ordinary microphones. This technology uses sound features as well as the direction from which the sound arrives (spatial information). For example, a feature of the sound of a strike at the goal is "sharpness of onset". This software processing makes it possible to clearly extract the competition sounds. The sound pick-up range is sufficient to cover the penalty area of a soccer field if the microphones are placed behind the goal.

2.Details

Previous technology for pin-point extraction of distant sound sources includes the 'zoom-in microphone' reported by NTT in April 2014.¹ That technology uses a sound pick-up system that consists of 100 microphones and a number of parabolic reflectors to achieve a confirmed precise directionality of within about 3 degrees. However, that system is physically large and was designed for permanent installation in a stadium or other such location.

In September 2014, NTT announced a microphone technology for high-noise environments that suppresses ambient noise in signals from multiple microphones.² With a few microphones, the size is not much different from conventional broadcast equipment, so actual field use is possible. That technology used mainly spatial information to estimate the frequency spectrum of the target sound and could distinguish the target sound from other sounds within a 60-degree range. However, in a soccer stadium for example, the sound of spectator cheering arrives from all directions, so it was not possible to clearly extract the sound of a ball kick, etc. with the processing as it was at that stage of development.

We therefore added processing that uses the characteristics of the sound of a goal shot (a sharp sound onset) to distinguish the target sound from the ambient noise (cheering), which was not possible when only spatial information was used. Using both the spatial information for the sound and the temporal properties of the sound source, we were able to extract the target sound clearly with two microphones.