There are three steps in automatic speech recognition (ASR): feature analysis, pattern classification, and language processing. Which of those three steps is the most challenging for a computer to perform? Why?
Which of those three steps is the least challenging for a computer to perform? Why? If ASR systems are to become automatic speech understanding systems, which step must undergo the greatest improvement in its capabilities? Why?