Gender recognition system consist of 2 parts.
1)pre-processing and feature extraction.
2) classification
using MFCC coeeffecients and also SVM.

Human face contains a variety of information for adaptive social interactions amongst people. In fact, individuals are able to process a face in a variety of ways to categorize it by its identity, along with a number of other demographic characteristics, such as gender, ethnicity, and age. In particular, recognizing human gender is important since people respond differently according to gender. In addition, a successful gender classification approach can boost the performance of many other applications, including person recognition and smart human-computer interfaces.

We have developed an algorithm for gender recognition based on AdaBoost algorithm. Boosting has been proposed to improve the accuracy of any given learning algorithm. In Boosting one generally creates a classifier with accuracy on the training set greater than an average performance, and then adds new component classifiers to form an ensemble whose joint decision rule has arbitrarily high accuracy on the training set. In such a case, we say that the classification performance has been “boosted”. In overview, the technique train successive component classifiers with a subset of the entire training data that is “most informative” given the current set of component classifiers. AdaBoost (Adaptive Boosting) is a typical instance of Boosting learning. In AdaBoost, each training pattern is assigned a weight that determines its probability of being selected for some individual component classifier. Generally, one initializes the weights across the training set to be uniform. In the learning process, if a training pattern has been accurately classified, then its chance of being used again in a subsequent component classifier is decreased; conversely, if the pattern is not accurately classified, then its chance of being used again is increased.

The code has been tested with Stanford Medical Student Face Database achieving an excellent recognition rate of 89.61% (200 female images and 200 male images, 90% used for training and 10% used for testing, hence there are 360 training images and 40 test images in total randomly selected and no overlap exists between the training and test images).