Any suggestions on how to increase the voice recognition accuracy in noisy environment?

Hi there,



please try below settings and see any improvement. thanks.



debian@beaglebone:~/usb_4_mic_array$ sudo python tuning.py STATNOISEONOFF

STATNOISEONOFF: 1

debian@beaglebone:~/usb_4_mic_array$ sudo python tuning.py MIN_NN 1.0

MIN_NN: 1.0

debian@beaglebone:~/usb_4_mic_array$ sudo python tuning.py GAMMA_NS 3.0

GAMMA_NS: 3.0



i use the 2k sine signal as background noise(first graph)

<LINK_TEXT text=“https://github.com/SeeedDocument/forum_ … e_tone.wav”>https://github.com/SeeedDocument/forum_doc/raw/master/reg/2k_sine_tone.wav</LINK_TEXT>



i record it with 1 channel firmware with STATNOISEONOFF = 0 setting as second graph.

<LINK_TEXT text=“https://github.com/SeeedDocument/forum_ … 2ksine.wav”>https://github.com/SeeedDocument/forum_doc/raw/master/reg/off_2ksine.wav</LINK_TEXT>



i record it with 1 channel firmware with above setting as third graph. it is much better.

<LINK_TEXT text=“https://github.com/SeeedDocument/forum_ … 2ksine.wav”>https://github.com/SeeedDocument/forum_doc/raw/master/reg/on_2ksine.wav</LINK_TEXT>