CM-PIE: Cross-modal perception for interactive-enhanced audio-visual video parsing