Sparse Canonicial Correlation Analysis for Mobile Media Recognition on the Cloud.
Yanjiang Wang, Bin Zhou, Weifeng Liu, Huimin Zhang · Journal of Multimedia · 2017
With the rapid development of the Internet technology and smartphone, people can easily capture and upload media information including text, audio, photos, and video. And then it becomes one critical demand to effectively and efficiently manage these personal multimedia that are often presented in multiple modalities. Canonical correlation analysis (CCA) has been widely employed for multi-modal data in many applications because of its promising performance in feature extraction and subspace learning for multivariate vectors. However, the traditional CCA may be difficult to interpret especially when the original variables are expected to involve only a few components. In this paper, we develop a mobile media recognition method on the cloud. Particularly, we propose sparse canonical correlation analysis (SCCA) on the cloud. SCCA can find a reasonable trade-off between statistical fidelity and interpretability. Furthermore, we employ a generalised power method to optimise the SCCA algorithm. Finally, we conduct extensive experiments for recognition on several popular databases including UCI datasets and USAA dataset. Experimental results demonstrate that the proposed SCCA algorithm outperforms the traditional CCA algorithm.