The structured time series (STS) classification problem requires the modeling of interweaved spatiotemporal dependency. most previous STS classification methods model the spatial and temporal dependencies independently. Due to the complexity of the STS data, we argue that a desirable STS classification method should be a holistic framework that can be made as adaptive and flexible as possible. This motivates us to design a deep neural network with such merits. Inspired by the dual-stream hypothesis in neural science, we propose a novel dual-stream framework for modeling the interweaved spatiotemporal dependency, and develop a convolutional neural network within this framework that aims to achieve high adaptability and flexibility in STS configurations from various diagonals, i.e., sequential order, dependency range and features. The proposed architecture is highly modularized and scalable, making it easy to be adapted to specific tasks. The effectiveness of our model is demonstrated through experiments on synthetic data as well as benchmark datasets for skeleton based activity recognition.