参数资料
| 型号： | AC300030 |
| 厂商： | Microchip Technology |
| 文件页数： | 35/64页 |
| 文件大小： | 0K |
| 描述： | KIT ACOUSTIC ACCESSORY |
| 标准包装： | 1 |
| 附件类型： | 配件套件 |
| 适用于相关产品： | 杂讯抑制/回声消除库 |
第1页第2页第3页第4页第5页第6页第7页第8页第9页第10页第11页第12页第13页第14页第15页第16页第17页第18页第19页第20页第21页第22页第23页第24页第25页第26页第27页第28页第29页第30页第31页第32页第33页第34页当前第35页第36页第37页第38页第39页第40页第41页第42页第43页第44页第45页第46页第47页第48页第49页第50页第51页第52页第53页第54页第55页第56页第57页第58页第59页第60页第61页第62页第63页第64页
Application Programming Interface (API)
4.3.1
Normalized Least Mean Square (NLMS) Adaptive Filter
NLMS is the fundamental adaptation algorithm used for estimating and canceling out
the acoustic echo. This filter tries to minimize the mean square error between the S IN
and R IN signals. Under conditions where there is no double talk, this will result in a set
of filter coefficients that approximate the acoustic path between the speaker and the
microphone. The filter thereby produces an echo estimate of R IN , which is then
subtracted from the S IN signal.
4.3.2
Nonlinear Processor (NLP)
The AEC algorithm by itself may not be capable of adequately modeling echo paths
that generate significant levels of nonlinear distortion. This necessitates the usage of a
Nonlinear Processor. The function of the NLP is to substantially suppress the residual
echo level which remains at the output of the NLMS adaptive filter, so that a very-low
returned echo level can be achieved even if the echo path is nonlinear. The NLP is
located in the send path between the output of the NLMS filter and the S OUT port of the
system. The NLP basically attenuates low-level signals (which are assumed to be
residual echo) and passes high-level signals (which are assumed to be desirable
near-end speech).
The AEC library offers two functions for using the NLP. The EC_applyNLP() function
applies the NLP action to the input buffer. The EC_setNLPLevel() function varies the
level of attenuation.
4.3.3
Howling Control
Howling is a typical problem in full-duplex communication. It builds up due to the
acoustic feedback path. One way to reduce howling is to shift the frequency of the
signal that is picked up by the microphone by 10 Hz to 20 Hz, before it is sent out over
a communication channel. This shift is usually not perceived as unnatural by the human
ear. The shifted signal appears at the destination loudspeakers and travels back to the
originator, shifted by another 10 Hz to 20 Hz. The signal travels many times through
this acoustic path and is quickly shifted out of the pass band, thereby reducing the
problem of unpleasant feedback.
The AEC library offers a function, EC_setHowlingControl() , to enable and disable
howling control.
4.3.4
Double Talk Detector (DTD )
Double talk is the condition that occurs as a result of two talkers on both sides (R IN and
S IN ) talking at the same time. During double talk, the signal S IN acts like uncorrelated
noise and may cause the coefficients of the NLMS adaptive filter to diverge, thereby
failing to effectively cancel the acoustic echo. To prevent such a condition, a DTD is
used to inhibit adaptation of the filter during periods of simultaneous far-end and
near-end speech. The DTD also inhibits the operation of the NLP to prevent loss of
near-end speech. In this algorithm, an energy-based double talk detector is used, in
which double talk is detected when Average Energy of S OUT > Average Energy of R IN .
The AEC library offers two functions, EC_setDoubleTalkHangover() and
EC_setAdaptionHangover() , to control the amount of DTD hangover. The double
talk hangover represents the number of frames after double talk has been detected for
which the AEC algorithm will wait before resuming application of NLP. For example, if
the hangover value is 6, the algorithm will wait for 6 frames before applying NLP again.
The adaptation hangover is controlled by EC_setAdaptionHangover() . The default
value for this function is ‘ 1 ’, so that adaptation resumes one frame after the end of
double talk.
? 2004-2011 Microchip Technology Inc.
DS70134F-page 35
相关PDF资料 |
PDF描述 |
|---|---|
| ACM-16 | CABLE MODULE FR3 16BIT 40-DIP |
| ACM-2X8 | ACTIVE CABLE MODULE DUAL 8 BIT |
| ACM-8 | CABLE MODULE FR3 8BIT 28/32-DIP |
| ACS9510EVB | EVALUATION BOARD FOR ACS9510 |
| ACS9550EVB | EVALUATION BOARD FOR ACS9550 |
相关代理商/技术参数 |
参数描述 |
|---|---|
| AC300031 | 制造商:Microchip Technology Inc 功能描述:ACOUSTIC ACCESSORY KIT SPEECH RECOGNITION - Boxed Product (Development Kits) 制造商:Microchip Technology Inc 功能描述:SPEECH RECOGNIZE KIT ACOUSTIC |
| AC3000-52 | 制造商:GE Energy (formerly Lineage Power) 功能描述:Front-End Power Supply 制造商:Lineage Power 功能描述:Front-End Power Supply |
| AC3000A | 制造商:Black Box Corporation 功能描述:HD VIEW TRANSMITTER |
| AC3000A-R2 | 制造商:Black Box Corporation 功能描述:HD VIEW TRANSMITTER |
| AC3000A-R2-W1 | 制造商:Black Box Corporation 功能描述:1 YEAR WARRANTY FOR AC3000A-R2 |
发布紧急采购，3分钟左右您将得到回复。