Data is plentiful these days. Open source software exists to do nearly anything you want. Artificial intelligence is the new panacea for everything from art to science. This leaves a question. It’s probably not the one you’re thinking of (Skynet is probably already online in more than one form). The question is how many of the problems we’re solving with these new techniques represent a step forward. I don’t think there’s an answer to this question that always applies.

So I’m going to write a series of posts to demonstrate an example of solving a solved problem. Hopefully, we can learn (together) some of the pitfalls and hopefully some of the benefits of attacking an old problem with new techniques. 

I’m going to take a basic digital signal processing problem, one that was solved when your parents got their first touch tone phone.

Can I build a system to decode DTMF (dual tone, multi-frequency)  signals? These are the tones that you hear when you press digits on your phone’s keypad. 

I’m going to solve it using a neural network. Then I’m going to demonstrate a common DSP technique (one that modern telephony systems still use). I actually don’t know what to expect, but I’ll do some validation and look for ways to break each approach. 

I’m going to touch on a lot of small but hopefully interesting “Data Science” things for people interested in some of the nitty-gritty.

OK, let’s get started!