Spiking neural networks (SNNs) encode information through discrete spike trains, which demonstrate unique advantages in brain-computer interface systems due to their biologically plausible spatiotemporal information processing mechanisms. Motor imagery electroencephalogram (EEG) signals are typical spatiotemporal data, and analyze EEG signals effectively is difficult because they naturally have a low signal-to-noise ratio. And substantial inter-subject variability. In this paper, we propose a lightweight spiking convolutional neural network model integrating discrete wavelet transform and efficient channel attention mechanism for EEG-based motor imagery classification, named WaveECANet. The model employs a three-stage progressive feature refinement framework and a residual-attention co-optimization strategy to enable multi-scale feature extraction and context-aware modeling. The WaveECANet can effectively capture both local and global spatiotemporal features while suppressing noise interference. Comprehensive experiments on the BCI Competition IV-2a dataset validate the proposed model, achieving a mean classification accuracy of 81.32%, outperforming other state-of-the-art motor imagery classification models. This spike driven method is more plausible for the development of EEG signal learning analysis and brain-computer interface technology.