To deal with the limitations of human action recognition systems that apply deep neural networks (DNNs) to 3D skeletal feature maps, we propose an improved set of quad features that enable better pattern discrimination when using a spectrally enriched circular convolutional neural network (CCNN). we compute the volumes of the time-varying quadrilaterals, by generating color-coded images, named spatiotemporal quad-joint relative volume feature maps (QjRVMs). Consequently, we introduce a new architecture CCNNs, which use cyclic multiresolution filters in a fourstream architecture, requiring only batch normalization and ReLU operations to identify multiple pixel patterns in a 102 class human action mocap data.