Gaussian mixture model (GMM for short) is an effective tool for data modeling and pattern classification. GMM assumes the data under modeling is generated via a probability density distribution which is the weighted sum of a set of Gaussian PDF. Via the use of EM (expectation maximization), we can identify the optimum set of parameters for GMM in an iterative manner.
Due to the flexibility of GMM, it has been successful applied to numerous applications of data modeling and pattern classification, including speaker/speech recognition, audio classification, image classification, background image removal, and so on.
The mathematical formula of GMM and the derivation its training method are a bit lengthy. Please refer to the PDF file for more details.
Data Clustering and Pattern Recognition (資料分群與樣式辨認)