当前位置:网站首页>Influencing factors of echo cancellation for smart speakers
Influencing factors of echo cancellation for smart speakers
2022-06-30 08:44:00 【Intelligent speaker design】
1、 What is echo cancellation
Echo cancellation (Acoustic Echo Cancellation,AEC), It refers to the estimation of echo signal size by adaptive method , This estimate is then subtracted from the received signal to cancel the echo .
2、 The purpose of echo cancellation
The purpose of echo cancellation is to work on the speaker of the speaker ( Play music or voice ) when , From the signal collected from the microphone , Remove the sound signal played by itself . otherwise , When the music is playing , Our sound signal will drown in the sound of music , Unable to continue effective voice control on the speaker .
3、 The principle of echo cancellation
The greater the similarity between the microphone signal and the recovery signal , The better the echo cancellation performance . Therefore, try to ensure the similarity between the microphone signal and the recovery signal , Try to reduce the influence of factors that affect the microphone signal quality .
4、 Influencing factors of echo cancellation
- The cut is
The amplitude of the stoping signal will cause the amplitude of the stoping signal to be greater than 0dBFS The signal of is truncated , The effect is that the signal played by the horn is completely distorted , The human voice is accompanied by strong noise . Microphone signal clipping will seriously affect the recognition effect , Therefore, clipping phenomenon must not occur in audio signal processing , Be sure to reduce the power amplifier or gain output if any .
Design principles :
The speaker plays a full sweep signal at the maximum volume , The loudspeaker shall not break 、 Resonance and other phenomena , The mic signal cannot be truncated . This is also one of the judging principles when specifying the output power of the horn .
- Harmonic distortion
Harmonic distortion means that the output signal has more harmonic components than the input signal . Harmonic distortion is caused by the fact that the system is not completely linear . There are many reasons for harmonic distortion , In the entire audio path , From the distortion curve of the loudspeaker unit , To the structural design of the rear cavity of the loudspeaker , Horn front net cover , Microphone receiving hole structure, etc , Will increase the distortion of the microphone signal . The more distorted the mic signal is , The worse the similarity between the microphone signal and the recovery signal , The worse the echo cancellation performance .
Design principles :
- Try not to use mini speakers 、 Ultra thin loudspeaker and other loudspeaker units with high resonance frequency , This kind of monomer has large low-frequency distortion ;
- If the rear cavity structure of the loudspeaker allows , Add passive radiator or guide tube design , Low frequency can be increased , At the same time, the nonlinear vibration of the horn is reduced ;
- Signal-to-noise ratio
Ambient noise and circuit noise will reduce the signal-to-noise ratio of the microphone signal , Especially in the case of low volume , The proportion of signal energy decreases , Increase of nonlinear distortion caused by bottom noise , The measured echo cancellation value is very poor . Generally, this noise refers to circuit noise , During the speaker test, the noise level of the test environment will be guaranteed .
Design principles :
- Select high sensitivity microphone , Improve signal receiving amplitude , Recommended values [email protected],1kHz And above ;
- Avoid excessive signal processing in the output circuit , Prevent the introduction of unnecessary signal crosstalk ;
- Automatic gain control
Automatic gain control (automatic gain control,AGC) An automatic control method for automatically adjusting the gain of the amplification circuit with the signal strength .
AGC is a kind of limiting output , It uses the effective combination of linear amplification and compression amplification to adjust the output signal of the hearing aid . When a weak signal is input , The linear amplifier circuit works , Ensure the strength of the output signal ; When the input signal reaches a certain strength , Start the compression and amplification circuit , Reduce the output amplitude . in other words ,AGC The function can automatically control the gain amplitude by changing the input / output compression ratio .
Design principles :
In the mining circuit ,AGC It shall be placed in front of the mining circuit , To ensure that the signal from the power amplifier to the recovery is consistent with the signal to the loudspeaker .
- Mechanical noise
In the category of intelligent projectors or robotic human devices ( Such as Tencent King glory doll speaker ) in , The variables introduced by fan noise and motor noise will cause the bottom noise of the microphone signal to be particularly large , The recognition effect is greatly affected . therefore , In this type of equipment , Pay special attention to the internal fan 、 The distance between the motor and the microphone board , If the structure space allows , The microphone plate can be sealed separately , Prevent internal transmission . - Circuit bad signal
When there are problems in circuit design , Usually, bad signals will be introduced into the recovery signal or microphone signal , Such as pulse 、 Clutter 、 Aliasing , Even lost data . In this case, the hardware circuit design should be modified first , Ensure that the circuit signal is clean , Then carry out acoustic test . - The DC offset is too large
DC offset (DC offset) It refers to the DC component in the AC signal , On the smart speaker , This phenomenon is usually caused by the bias voltage of the analog microphone .
Design principles :
Except for circuit design reasons , Using a digital microphone can effectively avoid this problem .

边栏推荐
- [kotlin collaboration process] complete the advanced kotlin collaboration process
- El input limit can only input numbers
- Detectron2 source code reading 2--- using the configurable decorator to build the dataloader
- Flink sql -- No factory implements ‘org. apache. flink. table. delegation. ExecutorFactory‘.
- 电流探头电路分析
- Introduction to MySQL basics day3 power node [Lao Du] class notes
- Advanced technology management -- how managers design and build echelons
- 2021-05-06
- VIM from dislike to dependence (21) -- cross file search
- 维基媒体基金会公布新商业产品“维基媒体企业”首批客户
猜你喜欢
![[untitled]](/img/45/368c41a74bf0738369c58c5963fca0.jpg)
[untitled]

【NVMe2.0b 14-3】Doorbell Buffer Config command、Device Self-test command

TiDB v6.0.0 (DMR) :缓存表初试丨TiDB Book Rush

Tidb v6.0.0 (DMR): initial test of cache table - tidb Book rush

vim 从嫌弃到依赖(21)——跨文件搜索

涂鸦Wi-Fi&BLE SoC开发幻彩灯带

swagger使用

Flink 数据偶尔数据积压导致checkpoint失败

【NVMe2.0b 14-2】Create/Delete Queue

电流探头的干扰源电流谱测试
随机推荐
技术管理进阶——管理者如何进行梯队设计及建设
Redis design and Implementation (III) | interaction between server and client (event IO model)
证券开户的优惠怎样才能得到?在线开户安全?
2021-04-29
JVM tuning related commands and explanations
Detectron2 source code reading 2--- using the configurable decorator to build the dataloader
2021-05-17
Redis设计与实现(八)| 事务
Interpretation of source code demand:a rotation equivariant detector for aerial object detection
TiDB v6.0.0 (DMR) :缓存表初试丨TiDB Book Rush
This point in JS
How can I get the discount for opening a securities account? Is online account opening safe?
Codeworks 5 questions per day (1700 for each) - the third day
Redis design and Implementation (I) | data structure & object
Swagger use
el-input 限制只能输数字
Circuit analysis of current probe
Flink Sql -- toAppendStream doesn‘t support consuming update and delete changes which
[untitled]
Advanced technology management -- how managers design and build echelons