More comments in the code would be great

[laurentopia]

I'm reading through SpeechRecognizer::spch_recg and would like to see a few comments such as
// flattening sound recording
// inject flat sound recording into the neural net
// NN result
etc... so we can learn from example.
thank you