SpeechCLIP. SpeechCLIP: Integrating Speech with Pre-Trained Vision and Language Model, Accepted to IEEE SLT 2022