How to get clean human voices in noisy environments (such as kitchen/living room/gym environments) is a problem that smart devices must solve when doing voice interaction and voice calls.
In real life, in a noisy environment, there are many kinds of noise, which also hinders smart devices from effectively acquiring human voices, these noises include:
• Sounds played by the device itself, such as TV/set-top box/Sound bar and other smart devices are playing music
• Steady-state and unsteady-state, divergent noise in the environment, such as fan/air conditioning and other basic environmental noise
• Point noise in the ambient space, such as noise such as the sound emitted by a TV fixed in a certain location
A high-performance voice interface is particularly important in such equipment. In addition to solving the noise problem, the P3610-2MIC can provide both long-distance pickup and voice interruption (Barge-in).
Such a front-end voice interface can output a clean and effective human voice for voice interaction (ASR) and conference calls (Communication).
Among smart devices, P3610-2MIC, as a cost-effective voice interface device, can be quickly applied to voice interaction (ASR) and voice call (Communication) use; In particular, it specially optimizes the front-end processing of ASR, which greatly improves the performance of voice interruption and speech recognition. Ideal for smart speakers, smart TVs, set-top boxes and smart gateways.
Overview of testing
The P3610-2MIC board voice performance test aims to introduce the basic performance parameters of the microphone for functional evaluation, including microphone sound pressure sensitivity, microphone consistency, DSP algorithm function test and other acoustic assembly test suggestions. Testing will use standard test equipment and software and will be conducted in a controlled environment.
The test conditions and tools are as follows:
1. Test equipment/tools: P3610-2MIC board, Type-C USB cable, hi-fi speaker or artificial mouth, Audacity/Adobe Audition audio software, test audio, decibel meter/sound pressure meter;
2. Test environment: The test should be carried out in the standard anechoic room/semi-anechoic room or conference room to avoid the influence of external environmental noise on the test results.
Test details
● Microphone sound pressure test
The AOP of a microphone is usually around 120 dB, and if the sound signal exceeds this value, it will cause the microphone to be intercepted, resulting in sound distortion.
Test method: Use a decibel meter close to the microphone, and the test ensures that the detected sound is within 120dB.
● Microphone conformance test
Each microphone pickup signal needs to maintain high consistency, and when assembling the structural housing, it is necessary to record all microphone raw signals uniformly and analyze and de-consistent, focusing on amplitude and spectral characteristics.
The consistency requirements of each microphone signal are the basic requirements of the entire XMOS array project before the array microphone can perform as it should.
Test method:
(1) Use a hi-fi speaker/artificial mouth to play a piece of white noise test audio;
(2) The P3610-2MIC board to be tested is about 30cm away from the artificial mouth for recording;
(3) Use Audacity/Adobe Audition to analyze the frequency of the recorded audio.
● DSP test
The following test content is all USB firmware test examples:
AEC test
The reference signal is one of the conditions that determine the performance effect of AEC, and the two reference methods currently used in the array scheme (USB/I2S), no matter which method is used, you first need to ensure that the device playback speaker can not have distortion, overload, and the speaker output and the digital I2S signal spectrum should be relatively matched.
Reference signal design requires the following considerations:
1. When using the USB path as the reference signal for pick-back, the I2S out back-end analog design uses a linear analog design to meet the reference signal and the I2S signal matching, do not use the EQ module chip, the EQ back-end analog output will not match the actual original reference signal spectrum information.
2. When using I2S as a reference signal, it is recommended to add an ADC chip from the back end of the equipment amplifier for voltage division sampling to ensure that the actual playback sound matches through I2S.
3. The loudness of the sound playback volume of the device at any time should not exceed the maximum sound pressure level that the microphone can withstand, and it is recommended that the maximum playback volume of the speaker reach the microphone sound pressure ≤ 120dB.
Test method:
(1) XMOS plays sound, pay attention to the shock absorption of MIC and speakers, and the tightness of MIC
(2) The sound source is about 65dB sound, and the distance from MIC pickup is 3-5 meters to test its performance
(3) The noise played, white noise is the best
(4) It should be noted whether the recovery method is USB or I2S, and the firmware used is different
IC testing
IC (Interference Cancellation, interference source elimination), the role of interference source elimination is mainly to eliminate the interference sound signal emitted by a fixed location point in the environment, which is regarded as an interference signal is eliminated by P3610-2MIC.
The IC algorithm does not need to access the reference signal, but there will be a convergence time process, the position of the interference source signal and the position of P3610-2MIC are relatively fixed, and the convergence elimination can be achieved after VAD vocal detection.
Test method:
(1) Use hi-fi speakers/artificial mouths to play noise test audio similar to washing machines and radios in anechoic rooms or conference rooms;
(2) Place the P3610-2MIC board in front of the hi-fi speaker/artificial mouth to record through USB; After a period of time, a vocal audio is played at the same time;
(3) After P3610-2MIC detection, use Audacity to view the processed audio.
VAD test
VAD (Voice Activity Detection). Controls the adaptability of the IC and AGC to optimize the output of near-end voice.
Test method:
(1) Use hi-fi speakers/artificial mouths to play a test audio with human voice and background noise in the anechoic room or conference room;
(2) Place the P3610-2MIC board in front of the hi-fi speaker/artificial mouth to record through USB;
(3) After P3610-2MIC detection and identification, use Audacity to view the processed audio.
NS test
NS (Noise Suppression), the role of noise suppression is mainly to suppress noise sources from whose frequency characteristics do not change instantaneously with time. These include scattered background noise and fixed noise sources, similar to steady-state noise and unsteady noise.
Test method:
(1) Use a hi-fi speaker/artificial mouth to play background sound test audio in the anechoic chamber;
(2) Place the P3610-2MIC board in front of the hi-fi speaker/artificial mouth to record through USB;
(3) After P3610-2MIC noise reduction, use Audacity to view and record the audio after high background_playback sound processing
AGC test
AGC (Automatic Gain Control), the function of automatic gain is mainly to dynamically adjust the audio gain, or set a fixed gain, so that the sound maintains the desired output level.
Test method: You can use Audacity software to record before and after, and compare the amplitude of the audio before and after
Fixed gain
Set to a fixed gain, the amplitude of the output audio signal will be amplified or reduced with the gain setting, but it should be noted that excessive gain will cause audio distortion, so it needs to be set reasonably according to the specific situation
Adaptive gain
Set to adaptive gain, it means that the gain value of the AGC is constantly changing, and the expectation is that the final audio signal will remain flat and the sound volume will be at the same level.
Acoustic structure guided testing
● Tightness test
The sealing is mainly to ensure that the pickup hole of the microphone only obtains the sound from the outside world in the pickup cavity hole of the structure shell, so as to prevent the sound from entering the structure and reflecting back and forth, resulting in serious reverberation.
In hardware design, it is necessary to consider the design of the tightness of the microphone in advance, usually the product is to increase the sealing sleeve to achieve sealing performance, and the hardware needs to reserve the position of the sealing sleeve to meet the structural acoustic requirements of the product.
Test method: Audacity recording to obtain each MIC raw data, in a recording test, the device locally plays white noise audio, use the sealing silicone sleeve to plug each MIC hole (pickup port) test, look at the difference before and after the plugging of the hole, in the case of good sealing, the difference before and after the plugging hole to maintain 30db or higher than 30db can be judged to be well sealed.
● Shock test
Shock absorption is mainly for the vibration generated by the sound cavity when the speaker plays sound without conducting to the microphone plate, so as to avoid the sound being transmitted to the microphone through vibration, and then avoid affecting the pickup of the microphone.
Regarding shock absorbing, the following suggestions are made:
1. In the design of the speaker cavity and microphone has sufficient distance, and there is absolute isolation;
2. Add shock cushioning materials, such as rubber pads, silicone pads, foam, and foam, to the connection between the product equipment body and the microphone module board.
About Papaya Technology
Papaya Electronics is deeply engaged in Hifi audio, microphone array and Alot and other fields, providing complete and reliable customized audio solutions, committed to creating value for customers.
Our vision: to become a world-class audio solution provider
Our mission: to make high-end sound quality no longer expensive, and help the development of artificial intelligence industry
Our values: focused, professional, responsible, open, win-win, empowering
Our development culture: to create a platform where everyone is an entrepreneur and empower employees to achieve win-win development