Copyright © 2003-2015 ATC Labs. All Rights Reserved.
Search ATC Labs
Licensing / Support
For details on implementations on other platforms please contact us at email@example.com
Virtually, all real time audio recordings suffer from distortion due to presence of noise of one form or other. A very common day-to-day example includes hum or a background noise while listening to a radio station or even in a conversion on a mobile phone. Technically, noise reduction modules involve complex algorithmic designs and are challenging considering the wide band ubiquitous nature of background noises.
Our technology addresses some of the essential requirements of real time noise removal methods like automatic removal of noise with no/least user intervention. Such self training technology is ideally suited in broadcast scenarios where repeated manual intervention for noise filtering becomes a nuisance.
Automatic Noise Removal
Detecting the presence of signal amid low Signal to Noise ratio (SNR) condition is always a challenge. Considering the fact that there are noise-like-signals even in a valid speech frame, ex. Fricatives, nevertheless makes the challenge tougher even under better SNR conditions. Some of our audio products like Audio Denoizer have a custom built Voice / Signal activity detectors to detect and train the noise statistics with high accuracy in real time even under adverse conditions. With the help of such detectors the noise removal algorithm quickly learns and adapts to varying noisy conditions thereby perform an efficient noise filtering of audio.
Audio noise removal algorithms inevitably change / distort the audio from its original composure, mostly for good, in removing noise. But some of the popular audio de noisers available in the market invariably distort the main audio in the pursuit of noise removal. Such a side effect of original signal distortion while applying a high degree of noise removal mostly leaves bizarre artifacts in the so called noise free audio.
Using our expertise in perceptual audio coding we incorporate state-of-the-art signal processing techniques to weigh and preserve the main signal characteristics of the audio even before any noise removal is applied. This makes sure that even under very poor SNR the algorithm does not compromise in distorting the main audio while removing noise.
In addition to removing noise in the audio our custom made tools improve the listening experience of audio. Multi-band Temporal Envelope Enhancement Processor (TEEP) is a signal enhancement approach to reduce the impact of noise. In particular it attempts to enhance the temporal envelope coherence between the high frequency and low frequency signal components making the audio more stable and focused.
Harinarayanan E.V, Deepen Sinha, Shamail Saeed and Anibal Ferreira “A Novel Automatic Noise Removal Technique for Audio and Speech Signals”, in the preprints of 123rd Convention 2007 Oct 5-8, New York, USA...Pdf