Skip to main content

iZotope's adaptive noise reduction

When it comes to audio, noise is the enemy. Broadcasters, however, face a more elusive enemy than many because noise can come from many sources and can vary wildly from source to source. Because broadcasting is often global and involves feeds from remote locations, environmental broadband noise such as lighting, traffic, generators, HVAC systems and crowd leakage can interfere with the content being produced. Combined with system-induced noise such as ISDN artifacts, telephone noise, bad wiring or poor gain staging, an otherwise great segment can be rendered useless. With these challenges, broadcasters struggle to provide consistent and intelligible audio.

Overcoming noise effectively in a live broadcast application is not only a matter of finding a specific type or source and suppressing it, but also having the ability to intelligently identify and adapt to new incoming noise in real time. This is where the iZotope ANR-B comes in. Gaining use in TV, radio and live sound reinforcement applications, it is a single-space rack unit featuring two-channel (or linked stereo) operation in two modes: Adapt and Manual. Adapt mode identifies incoming noise completely automatically and in real time. This mode works well for suppressing steady-state noise that is unpredictable or that changes over time. Manual mode allows for the capture and suppression of a specific noise profile, which can then be stored as a preset for recall at a later date. This mode works well in a studio when specific system noises need to be suppressed. In both modes, the amount of suppression is easily adjusted by a single knob, which lets the engineer focus on other high-level tasks; this is critical in any live application.

At the heart of the unit is the company's adaptive noise reduction (ANR) algorithm, which is designed to suppress unwanted background noise in an audio signal, even if the background noise is changing over time. (See Figure 1.) Its application in broadcast workflows calls for automatic operation, so the hardware needed to have a “set it and forget it” user interface.

When presented with an audio signal, the ANR-B uses a filter bank to divide the audio into hundreds of frequency bands. After the filter bank, the noise reduction technology works to identify the speech signal and steady-state noise sources. The noise reduction intelligent identification system is the key to its fully automatic operation. Once the signals are identified, the unit uses techniques found in high-quality audio restoration processors to suppress background noise and isolate the voice as transparently as possible. One basis for these techniques can be thought of as a massively multiband gate, working in each frequency band to suppress noise but allowing desirable signals to pass through unaltered.

The key to the unit's success is a careful balance between high-quality processing, fully automatic operation and low-latency performance. This requires high-power processors (one Analog Devices SHARC DSP for each channel), as well as extensive testing with several customers on countless hours of audio. Most noise reduction systems require an experienced operator who needs to constantly monitor and adjust parameters. With the ANR-B, the signal monitoring and parameter adjustments are left to the unit, allowing the operator to focus on the creative decisions. While the system is designed for automatic operation with one knob, it provides essential metering information with meters for input, output and noise suppression amount, as well as a momentary residual noise switch for monitoring removed signal.

Bruce Bartone is director of pro audio sales at iZotope.