Improve voice intelligibility with Active Audio's speech Conformer

Active Audio claims that its new speech Conformer algorithm can dramatically improve intelligibility, using a technique that has never heard before. It is now possible to have a clear sound whoever speaks into the microphone, says the company.
Author:
Publish date:
2012.jpg

Active Audio claims that its new speech Conformer algorithm can dramatically improve intelligibility, using a technique that has never heard before. It is now possible to have a clear sound whoever speaks into the microphone, says the company.

The intelligibility of voices carried by a PA system depends on several factors: the quality of the PA system, the acoustics of the room, background noise, and also the voice itself. In a given environment, one voice might be perfectly intelligible over the PA system, while another voice is not. This is a familiar experience on public transport, in conferences or in churches.

The quality of the speech signal depends on the voice of the speaker, and on the way they speak in the microphone – in particular the position (angle and distance) and movement of the speaker relative to the microphone. This results in variations in the level and timbre of the signal. Variations in level can be limited by using a compressor, which attenuates very loud voices (or loud parts in a speech) and boosts soft ones.

In addition to using a compressor, Active Audio’s speech Conformer is an adaptive algorithm that corrects the timbre of voices in order to enhance their intelligibility. The Conformer analyses the spectrum of the input signal and applies the necessary gain so that the spectrum of the output signal meets the target spectrum chosen by the user. In contrast to a simple equaliser, the Conformer is adaptive: it will correct a voice whose spectrum is different from the target spectrum, but not a voice with a spectrum close to the target.

In the interface of the speech Conformer, the user can:
- Choose the target spectrum, determined by the level in bands Warmth, Presence and Brightness
- Specify the maximum gain the Conformer is allowed in each band
- Launch a test procedure to ensure the Conformer does not produce howlround – when a voice requires a high gain in a given
band
- View the gains in each band in real time. When changing between speakers, the adaptation of Conformer gains is typically 5s, and very stable (about +/-1.5 dB).

The Conformer is embedded in the Speech Processor application for Active Audio’s NUT processor. This also features an automixer, a feedback killer, as well as more conventional algorithms for compressor, EQ, delay and router. In this application, the Conformer memorises the gains for all channels. Thus, when a speaker returns to the microphone after handing over to someone else, the Conformer reloads the gains it had beforehand – adaptation is immediate.

“The Speech Processor application is online now, and many of our clients are eagerly waiting for it after the presentation we made on our booth at Prolight & Sound,” said Xavier Meynial, technical director of Active Audio.

www.activeaudio.fr

Related