Reliable delivery of 3D video contents to a wide set of users is expected to be the next big revolution in multimedia ap- plications provided that it is possible to grant a certain level of Quality-of-Experience (QoE) to the end user. During the last years, several cross-layer solutions have proved to be ex- tremely effective in tuning the transmission parameters at the different layers of the protocol stack and in maximiz- ing the perceptual quality of the reconstructed 3D scene. Among these, Cognitive Source Coding (CSC) schemes (de- fined in analogy with Cognitive Radio systems) make possi- ble to improve the quality of the 3D QoE at the receiver by adapting the source coding strategy according to the state of the transmission channel and to the characteristics of the coded signal. This knowledge also permits an optimization of the computational complexity required at the encoder. The paper presents a CSC architecture that analyzes the 3D scene, identifies the different elements, and chooses the most appropriate coding strategy via a classification of the features of each element based on Support Vector Machine theory. Experimental results show that the proposed ap- proach permits improving the quality of the received 3D signal with respect to traditional cross-layer techniques and reducing the computational complexity of coding operation.