About Noise Suppression (NS)

Noise Suppression technology reduces stationary and transient noises in single-channel speech signals increasing the signal-to-noise ratio, improving speech intelligibility and reducing listening fatigue.

Similar to most competitive noise reduction techniques, Alango Noise Suppression technology operates on a single channel PCM signal. It divides the input signal on a number of frequency subbands, estimates ambient noise and signal levels in each subband and then passes or attenuates the subband signals according to instant signal-to-noise ratios. The output signal is then recombined from the processed subband signals. Alango NS technology is distinguished by its ability for fast and precise tracking of ambient noise levels while preserving very high output speech quality. This allows suppressing both stationary and transient noises such as passing cars.

There are two versions of Alango Noise Suppression technology: “High Resolution” (HR) and “Low Delay” (LD).

High Resolution utilizes Alango patent pending dual stage frequency subband decomposition scheme providing high frequency resolution for processing while introducing relatively small additional delay into the subband analysis/synthesis scheme. High Resolution version delivers uncompromising output speech quality even for bad initial signal-to-noise ratios with only relatively modest computational requirements.

Low Delay version uses different frequency subband decomposition scheme also having very low data memory requirements. Sacrificing frequency resolution at high frequencies, it introduces only a negligible processing delay (< 3 ms). Since its almost not existing processing delay, Low Delay version is the technology of choice for time delay critical applications such as live sound reinforcement systems. It has also being licensed for noise canceling on speaker channel in wired, amplified phones where delay is unacceptable due to the side tone.

Noise suppression mode (fixed or dynamic), amount of noise suppression and noise adaptation time are configurable parameters. In the fixed mode noise is always reduced (SNR improved) by a specified number of times. In the dynamic mode the amount of noise reduction depends on the noise level. The more noise is present, the more it is reduced allowing to clean transient noises such as passing cars.

Technical information

HR version:

Noise suppression up to -30 dB;
Processing delay: < 30 ms;
Supported sampling rates: 8 KHz, 16 KHz;
Noise adaptation time: 150 - 1500 ms;
Available on: Blackfin Processor (Analog Devices), ARM11, ARM9, ARM7, Kalimba DSP (CSR).

LD version:

Noise suppression up to -15 dB;
Processing delay: < 3 ms;
Data memory: 350 words;
Noise adaptation time: 150 - 1500 ms;
Available on: Blackfin Processor (Analog Devices), TMS 54xx (TI), Teak Lite (CEVA-DSP).

Alango NS fully complies with VDA specification.

Interactive audio examples

Use control buttons on the interactive screenshots to play/stop/navigate.

Audio example 1 (stationary noise suppression mode)

Unprocessed original microphone signal containing speech and background noise:

Processed signal after noise reduction by 15 dB (fixed mode):

Audio example 2 (changing noise suppression mode)

Unprocessed original microphone signal containing speech and background noise of cars passing by:

Processed signal after dynamic noise reduction:

NS and other technologies

Although it can be used standalone, Noise Suppression technology is also integrated with other Alango technologies into one Voice Communication Package (VCP). The package also includes acoustic echo cancellation, dynamic range compressor, speech enhancement, (optional) adaptive dual microphone and several others.

Additional information

NS leaflet >>

Please contact us for more information, audio examples and real time demo availabilities.

go back to the Technologies page.

Site development: http://radzishevsky.info, 2006-8