This is an ambitious project. I definitely think it is possible, but you will need time.
Hardware-wise an easy solution would be the Audioinjector soundcard, which sports a build-in microphone .
Software-wise, I would encourage you to check out the free speech recognition toolkit Kaldi .
It can be trained to recognize audible sounds.
You will first need to record data and annotate it.
Then, you can use the annotated samples to train and test the recognition system.