What microphone did you use for raspberry pi ?

[dingusagar]

The detection is decently working when run on the computer. In raspberry pi, since there is no audio-in port, the only option is usb microphone. I tried this cheap usb sound card, but the noise is too high that the baby-cry is not being detected.

If anyone has tried this project on raspberry pi, please tell me which microphone you used for recording.

[giulbia]

Hi, I had the same problem but with another cheap usb microphone (you can find it on amazon as well).

What worked eventually is the Blue Yeti. So it served to prove the concept but it has some drawbacks:

• the price: it doesn't really make sense to use a Raspberry Pi which costs about 30$ if you have to invest 200$+ for the mic
• the shape: it's quite big to put it in a crib.

If anyone knows of a cheap, small, good quality usb mic it'd be good to know!

[dingusagar]

I tried using the sox software to remove the noise after recording. But it is still not detecting.

I had another plan. Not sure whether it will work out.

Record using the cheap usb microphone to get the noise alone. Now merge this noise with every audio file in the data set. and train again. What do you think about this approach. Will the SVM model be able to differentiate the classes based on the 5-6 features used in feature extraction.

[giulbia]

I like the idea about recording using the cheap usb microphone to get the noise alone. But rather than adding it to the training set I'd rather subtract it from the recorded signals at prediction. It would be an additional step before feature engineering only at prediction. Do you see what I mean?

[dingusagar]

Yeah....I will try that Thanks