Evaluation of incremental learning algorithms for HMM in the recognition of alphanumeric characters

Hidden Markov models based dynamic hand gesture recognition with incremental learning method

“…An incremental learning method is promising to overcome the shortcomings found in traditional machine learning approaches. Incremental learning method consists of techniques to enable classifiers to gather more information from unseen data but do not forget old information, without having to access previously learned data [18].…”

“…The incremental learning method we applied in our system is ensemble training (ET) and ensemble learning (EL) [18]. ET consists of independently do the learning of each of the observation sequences from the training set so that each sequence generates an HMM.…”

Diverse videos synthesis using manifold‐based parametric motion model for facial understanding

“…Adding a number of artificial samples to training sets is an effective method of improving the predictive capability of PIRS, including virtual data generation approaches, which are frequently used in pattern recognition, where prior knowledge obtained about the particular subject is used to create virtual samples. These virtual data points can be used to improve generalisation [10]. Niyogi says, ‘In certain function learning contexts, the framework of virtual examples is equivalent to imposing prior knowledge as a regulariser’ [11].…”

Video‐based facial expression recognition by removing the style variations

“…Synthesis methods have been proposed for the generation of virtual samples of various poses and expressions from a given face image [8,9]. Virtual data were used to improve the performance person-independent systems [10]. In summary, the novel contributions made by this paper are as follows: † Employing different mappings for discriminant analysis, dimensionality reduction and style transfer in order to increase the generalisation ability of the recognition system based on dynamic modelling of facial expression.…”