This is the respond from Imilab when I sent an email:
Dear
Good day !
First l am really sorry for the camera trouble ,
This camera is not very sensitive for checking baby crying.
In the case of continuous crying, the better the collection effect, for example, intermittent crying has a greater probability of not being recognized. Need to continuously collect 30s and above to achieve better recognition results. The listening position recognition function is still in the stage of optimization and perfection, and Ai also needs continuous learning