R262 datasheet(13/28 Pages) ONSEMI

13 / 28 page

BelaSigna R262

http://onsemi.com

Advanced Speech and Noise Management

System

BelaSigna R262 contains an advanced noise reduction

algorithm inside its ROM memory which preserves voice

quality while significantly reducing background noise. The

algorithm identifies and enhances human speech. Many

two−microphone algorithms depend on beamforming and

rely on the speech originating from a specific direction. As

a result, the performance of these algorithms can vary

greatly depending on the location of the speech and the noise

relative to the microphones, and typically will distort the

speech as well as the noise. Other algorithms that depend on a

speech−facing microphone and a noise−facing microphone

can require a lot of algorithm tuning for reasonable

performance. BelaSigna R262 is different as it is a true 360

self−adaptive multi−beamformer that distorts neither speech

nor noise. It applies attenuation or gain based on intelligent

classification of signals, resulting in a signal where the noise

is attenuated, and the speech is enhanced. The attenuated

noise is left as natural as possible so other downstream

processing is still effective. At the same time it performs

de−reverberation of the received signals. The solution is

ideal for any voice capture device requiring consistent

performance in noisy environments. Unlike other purely

directional solutions, BelaSigna R262 will not drop your

voice when you change orientation of the handheld device.

The algorithm works with two forward−facing microphones

spaced 10 to 30 mm apart (typically), and no tuning or

matching of microphones is required. The speech source can

come from any direction (normally in front of the

microphones), as long as it is within the working sphere of

the current algorithm mode. The algorithm mode is set based

on the required talking distance (distance between the mouth

reference point and the microphones), with noise reduction

performance decreasing as talk distance increases.

BelaSigna R262 can be set to perform maximum noise

reduction with a talk distance of 5 to 10 cm, mild noise

reduction with improved voice pickup at up to 5 m away, or

anywhere in between.

The noise reduction algorithm built into BelaSigna R262

can provide two processed outputs, each with different

tuning or operating parameters, which can serve two

different use cases simultaneously. An unlimited number of

use cases can be obtained by configuring BelaSigna R262,

where the user can easily select the appropriate distance

between the device’s microphones and the user’s mouth.

Any two of these use cases can be selected to the output

stage, allowing the target application to simultaneously

support use cases with mouth−to−microphone distances

varying between 5 cm and 5 m.

For additional details on the configuration of the

algorithm, please refer to BelaSigna R262 Communications

and Configuration Guide.

Algorithm Performance

BelaSigna R262 offers three main algorithm modes; the

close−talk mode aggressively filters noise and manages gain

to pick up speech within 5 to 10 cm from the microphone

array. The algorithm is not sensitive to handset holding style

in any mode, including close−talk mode. The near−talk

mode can effectively pick up speech within 0.5 m to 1 m

from the microphone array, an intermediate situation

between far−talk and close−talk modes. This is suitable for

capturing speech when using the device at arm’s length. The

far−talk mode detects and enhances speech from a distance,

but removes less overall environmental noise compared to

close−talk mode. Far−talk mode is suitable for recording

lectures or processing conference calls on speakerphone. In

addition to these three main modes (close, near and far),

BelaSigna R262 features a mixing capability, which allows

the selection of any two of these three main modes combined

through a mixing ratio to create any intermediate mode,

allowing manufacturers to fine tune the algorithm based on

the microphone−to−user distance on the target device.

A selection of performance metrics are shown in Table 7

for some input signal−to−noise ratios and noise types.

Signal−to−Noise Ratio Improvement (SNR−I) was

measured according to the ITU−T G.160 standard, with

BelaSigna R262 operating on reference hardware, in

conjunction with common omni−directional microphones

(ECMs) positioned 11 mm away from each other. To verify

that subjective quality was maintained through the noise

management process, a Perceptual Enhancement of Speech

Quality (PESQ) measurement was also taken for each

condition. PESQ improvement correlates to MOS

improvement, the widely−accepted subjective standard in

voice quality.

R262 Datasheet(PDF) 13 Page - ON Semiconductor

R262 Datasheet(HTML) 13 Page - ON Semiconductor

Html Pages

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28

Datasheet Download

Link URL