Do self-supervised speech and language models extract similar representations as human brain?