Ask HN: How do I go about audio background noise removal? | Hacker News Reader

Ask HN: How do I go about audio background noise removal?

2 pointslovelearning3 comments

I'm looking for suggestions from you folks on the following audio processing problem: After recording a speech, is it possible to remove background environmental noise (mostly vehicle engine and horning noise) from it?

Can I isolate just the voice somehow from a single sample?

Or can I isolate the voice by using an additional microphone placed outside to record the external noise, and then somehow "subtracting" the noise recording from the voice?

I'm more interested in understanding audio processing from first principles, understanding the math, reading relevant papers, exploring signal processing algorithms and implementing them, than in buying suitable microphones or using available software.

What approaches do you suggest and what topics do you think I should explore?

I would suggest you start with reading "Signal Processing and Linear Systems" by Lathi (great book, had it on a course) so that you gain some understanding on signal processing.

After understanding filters, and frequency analysis, you will be able to ask more specific questions than "how to isolate a voice from a single sample?". I am just finishing a course on Image Processing and while some of the stuff that I learned on Signal Processing was useful, I would describe it as being only primitives (think of as integers, doubles, and control structures in programming languages) and we explore higher level of abstractions (segmentation and classification techniques). I assume the same is true for audio signals, and recommend as well to read a specialized book on the subject. I am sorry I cannot recommend one.

Thanks for the suggestions, really appreciated.

I'm not an expert, but I was during last 3 years editing the voice for videos to upload them to YouTube.

The best improvement was to buy a headset with microphone (not very expensive one ~$30). It killed most of the background sound and the mmmm. Also the volume voice is more stable, because the distance to the mouth is constant. I don't know if this is applicable to your problem, but a little hardware helps a lot.

Reply on news.ycombinator.com