From ftm
MnM is a set of Max/MSP externals based on FTM providing a unified framework for various techniques of classification, recognition and mapping for motion capture data, sound and music.
Features
- Hidden Markov Models
- Principal Components Analysis
- Singular Value Decomposition, LU and QR decompositions
- Non-negative Matrix Factorization and sparse decomposition
- multi-dimensionnal M to N mapping based on examples
- Multi-dimensioannal autocorrelation
- Matrix/Vector Statistics (min, max, mean, std, histogram, mahalanobis distance)
Publications
- NIME 2005 paper on MnM (PDF)
The MnM package is released within the FTM distribution..