Amazon Echo - Ask for an algorithm to avoid the background command

Question

Amazon Echo - the algorithm to avoid the ambiguous background command from the voice record being played:

This is a hypothetical question based on the principle that it could happen: If one uses Amazon Echo to play a voice recording or a video which contains some sentences like "Alexa, could you ... , " What will happen?

1. If the video/recording contains a sentence like: "Alexa, could you stop the video?" What will happen?

2. If the video/recording contains a sentence like: "Alexa, please increase the volume to 8?" Meanwhile, you command to the Echo that "Alexa, please decrease the volume to 4?" Could it distinguish which one is the command to fulfill?

Would Amazon Echo be able to neglect the voice recording or the video being played, and not to misunderstand it as a real command from the real human? What kind of algorithm is designed for the Amazon Echo program to deal with this situation?

Answer 1

When a device plays noise and has a microphone, then the manufacturer uses a digital signal processing technique called "echo cancellation" to subtract the noise it makes from the sound picked up from the microphone. This includes phones, headsets, your computer (skype does echo cancellation in software), and the Amazon Echo.

Because of echo cancellation, the Amazon Echo cannot hear itself, so it won't respond to commands that come out of its speakers.

There are lots of questions about echo cancellation on SO, which you can easily find now that you know the magic words. The algorithms are too complex for an SO answer, but you can generally get an open source implementation for whatever environment you're working in.

That's the real answer... But your post reminds me of a book called "Goedel, Escher, Bach", by Douglas Hofstadter. He discusses a similar question about record players, which can pick up their own sounds.

That question has a super-interesting answer: http://genius.com/Douglas-hofstadter-contracrostipunctus-annotated