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