I am recording from the Mic array but the volume is much lower than other microphones on my system. As a result I get better speech recognition from an ancient web cam microphone than the Respeaker. Is there a way to boost the volume ? I have tried changing the AGCGAIN (AGC is turned on). Initially AGCGAIN was 8.07027126849. I set it to 16, 32 and finally 1000, but there did not seem to be any effect on recorded volume.
Edit:
Further experimenting with changing the AGCGAIN showed that it does indeed change the recording volume, but only temporarily. The effect seems to persist until the Respeaker hears something ‘significant’, and then AGCGAIN is modified. In a quiet environment the value I set until the first time that I speak. I assume that it is automatically set to what the system believes is a good value. So is there a gain setting that biases the AGCGAIN to be higher ?
Hi,
Did you use 4 channels for recording? seems like 2 channel!
The firmware that I used is 1_channel_firmware_12.06dB.bin. I only seem to get one channel from this. How would I use the 6 channel firmware for speech recognition ?
I am connecting to Respeaker through Python SpeechRecognition package (using recognize_google method). I also use Respeaker when using Google speech recognition from Chrome web browser. If I am doing something wrong please let me know.
if you are using Respeaker 4 Mic array, you can use 4 channel to record the audio.
</s>arecord -Dac108 -f S32_LE -r 16000 -c 4 hello.wav // only support 4 channel
aplay hello.wav // Audio will come out via audio jack of Raspberry Pi <e>
for more details : ReSpeaker 4-Mic Array for Raspberry Pi Wiki
We can confirm that something is very fishy with the AGC.
Even though AGC is turned off by AGCONOFF 0 the AGC is still active. If you produce low sounds that barely makes the leds light up the gain will slowly increae to insane levels. The reported AGCGAIN does not correpsond to the actual AGCLEVEL ans manually setting the AGCGAIN will only affect the actual gain for a very short period, then it goes back to the high gain again. Making some louder noise makes the gain go back to normal levels again. This with the AGCONOFF set to 0 (it actually does not matter which value it has 0/1)?!?
The AGCMAXGAIN only seams to affect the displayed AGCGAIN, not the actual gain. Setting the AGCMAXGAIN to min value 1.0 does not prevent the gain to go to insane levels if the mic gets low sound levels (that barely lits the doides) for a while.
We have been using the 1-channel firmware since other software we are using are getting confused by the extra channels. Seeed, could we please get a working firmware or access to the source-code for the existing so that we can correct the fatal flaws that exists in the AGC algorithms?