I want to select the loudest digital audio signal (PCM-16bit) from various digital audio signals(PCM-16bit). I have tried to do it on a dB basis but unfortunately, there is hardware dependency that different mobile phones produce different dB levels based on their microphone. Is there any technique or algorithm to select the loudest stream other than dB comparison?
Though I am not pretty sure of whether this may work, I would suggest extracting other features from your audio signal (such as Mel Frequency Cepstral Coeffients (MFCC), Energy, Pitch, Zero Crossing Rate, Spectral Centroid etc. and so on). Refer to Audio Signal Processing related papers, which would give a good starting point towards this task. Hope this helps.
Joyjit Chatterjee , I have tried dB, RMS(Energy), Frequency, Pitch or Fundamental frequency, Zero crossing rate and even autocorrelation of signals. I have tried different algorithms in the time domain but it didn't provide what I expect. If you have any research paper, could you please share it with me?
I am not aware of any directly relevant papers in this area, but I believe that using relevant features can make all the difference. I am not sure why loudness isn't working well as I would have expected it to work in conjunction with an array of other features. Maybe you can try some thresholding algorithms to see whenever the loudness is above/below a particular level, you would be classifying it into a certain class? Hope this helps, and I would suggest you to look for papers in Audio Signal Processing and libraries like pyAudioAnalysis. I am sure that if you use pyAudioAnalysis in Python for instance, you will come across a suitable metric to achieve the task.
In my opinion you need some sort of common reference. How else would you know how loud an original analog signal has been if you don't know the mic gain of the mobile phone it was recorded with. In order words, if you want to compare these signals, you need to make them comparable/relatable in the first place.
I guess there will be no way around calibrating the mobile phone microphones.
René Erler Yeah.. I came with an algorithm to normalize minimum and maximum dB of different smartphones in a certain range and based on that normalized value I selected Loudest stream.
- Badges
- Science topic