In this paper, we present a comprehensive introduction about the modern ASR technology in terms of advanced techniques and applications as well as its challenge nowadays. We discuss the basics of Automatic Speech Recognition (ASR) systems such as acoustic modeling, language modelling and decoding algorithms. This work covers state-of-the-art techniques ranging from deep learning based models, attention mechanisms and transfer learning used in ASR. We then summarized some of the longer-standing issues in NLP, such as robustness to noise and low-resource languages or real-time processing. The paper also reviews the expanding use cases of speech recognition across industries such as voice assistants, automotive systems, healthcare and education. In the last part, we discuss societal implications for this technology and potential future research in speech recognition.