if we extract low level features and use them as classification attributes, they are large in number, so please suggest any attribute selection method which will reduce no. of attributes while still maintaining the accuracy of the classifier.
Dear Pooja Joshi
I do not know what kind of features , and which classifiers that you have used .
But ,
I recommend you to use Principal Component Analysis (PCA)
first for normalization through minimize the redundancies.
Second, for dimensionality reduction to reduce the number of features.
Regards,
In addition to the first answer, you can use mRMR feature selection method and some technics to select best features in your data.
You can find it in our publication:
Article Collection and Analysis of a Parkinson Speech Dataset With M...
PCA and mRMR mentioned above are two mature methods of feature reduction. If response time is a concern you can simply use LDA/Fisher discriminants.
on what basis you want to classify the audio or music files? For example- timbre classification -MFCC will be useful and so on. Accordingly you can select different low level features.
- Badges
- Science topic
- Similar topics
- Physical Sciences
- Waves