I am new to fmod and I love it very much. I am doing real time sound signal processing such as sound localisation and speech recognition. The background of recording is noisy. How should I use fmod to give me a clearer sound recording? Any bandpass filter or spectrum subtraction examples? I know there is a dsp moduler, but it looks like it apply dsp to speaker only rather than the recorded data. sorry, I may misunderstand in some points if I am wrong.
Thanks very much in advance.
- kingwinter asked 7 years ago
That’s a pretty involved question. On the FMOD API side, you definitely want to build a DSP chain, play your sound through it, and then write out/store/whatever the results. There are good examples of how to do all that stuff in the examples that come with the API (basically record sound, build DSP chain, play recorded sound through DSP chain, save result). You’ll have to put the pieces together yourself, but that shouldn’t be so hard.
As for how to do really good noise reduction on the human voice, well, that’s super tricky. Dolby Labs has been working on that problem for a few decades and their results are still only so good. Pretty sure much of that stuff is protected under pattents, but you could probably find out what the general state of the art is just by some google searching of white papers.
- audiodev answered 7 years ago
Please login first to submit.