09-13-2012 04:48 AM
Hi All,
My problem is the following, On one side I have to record the output of a synthetic human voice. One a second side I have a library of all the sentence that my human voice can say.
I need to be able to check if one or more pattern that I have in my library are recorded in my record.
Is there any way to do some kind of waveform pattern recognition in a larger waveform ?
Best regards,
09-13-2012 07:20 AM
First, consider extracting key features from your waveform (harmonics, kurtosis, octave analysis, wavelets, FFT, Cepstrum, etc). There are a number of these available in LabVIEW (Sound and Vibration Toolkit, Advanced Signal Processing Toolkit, LabVIEW advanced analysis library, etc.
Then consider pattern matching algorithims such as Statistical Pattern Recognition, Logistic Regression, etc. Consider the NI LabVIEW Machine Learning Toolkit, or the WatchDog Agent(R) Prognostics Toolkit for LabVIEW. Both of these tools can be found on the LabVIEW tools network.
Hope this helps you get started.