C2SER. We propose C2SER, a novel audio-language model designed to enhance the stability and accuracy of speech emotion recognition through contextual perception and chain of Thought (CoT).