Principles behind Speech Guard

The exciting advances in Oticon Agil's is made possible by the new signal processing system with Floating Linear Gain. Floating Linear Gain provides improved waveform fidelity, maximises speech audibility and offers ultra fast and comfortable loudness (transient) protection.

When the sound is steady with minimal changes in loudness, the gain is linear and amplification slowly adapts to changes in the input level. For sounds which are too soft or too loud, Oticon Agil responds rapidly with fast compression for an instant, to quickly return to an appropriate linear gain window. This assures that speech and spatial cues are amplified in a more natural and consistent way across different listening environments and this is the basic principle of Floating Linear Gain.

Speech Guard in action

Participating in conversation requires an ability to follow speech over time. This means being able to focus on the individual speaking while ignoring competing sounds such as sudden loud sounds (transients) or stable background noise. Speech is dynamic; its loudness level changes over time. Loudness changes are important natural cues that make it easier to differentiate one voice from another, making it easier to follow conversations.

This unprocessed speech envelope with a transient shows a waveform with a 6 dB modulation depth and no distortion in the signal. The transient is a short peak (50 ms) of 80 dB. With Floating Linear Gain the output signal preserves the 6 dB modulation depth and the distortion-free quality.

The transient is kept audible at a loud but comfortable level, where it does not disturb or attract attention. The Floating Linear Gain reacts so quickly when handling the transient that the ongoing signal is perceived as intact. With a conventional compression system the modulation depth is reduced and the envelope waveform is corrupted. When handling the transient the compressor attacks too slowly, making the transient very loud and annoying. It also releases too slowly, which makes the ensuing speech signal sound distorted and, in the worst case, inaudible.