Hi folks,
I hope that you are doing well. I am doing a research project in which I have to map the spectral features of sounds of audio of a fluid flowing through a pipe to the actual flow rate of the fluid through the pipe. I have been using various types of features in the past. For instance, I have used the MFCCs as features from the RASTA-MAT package. Moreover, I used different features that are basically statistics about the fourier spectrum from the audiofeatureextractor of Matlab. Notable features from here are pitch, spectral flatness and spectral roll off point etc.
Eventually, I map these features to the flow rate using a time series neural network. And so far I have been getting mixed results but not as nice as it could have been.
I wonder if I am missing something important, especially in terms of features.
I shall very much appreciate your feedback. And anyone who is willing to lend a serious hand of help, I shall also acknowledge him/her formally on my project in whatever way it is best possible in a formal sense.
Best regards,
Muhammad Adil Raja
Postdoctoral researcherDepartment of Computer Science and Information Systems (CSIS)
University of Limerick
Castletroy, Limerick
Ireland