Problem 2Speech Technology (50 points)(a) (10 points) Explain the use of Hidden Markov Models (HMMs) and n-gramsin a speech recognition system, what a r e their primary uses and how mightthey be trained.12(b) (15 points) In text-to-speech synthesis describe what the following stages inprocessing involve: “text analysis”, “linguistic processing” and “waveformgeneration”. Discuss some of the issues that must be addressed in thesecomponents, giving examples where appropriate.34(c) (25 points) You are asked to add speech to a robotic receptionist that sitsat the front desk of a large organization. The robot has an animated faceand can track users with a laser rangefinder.The system already has been used with type-in and print out information.Discuss what speech recognition and text-to-speech comp onents would beused in your design, and what their requirements and restrictions would be.How would these compo nents interact with the existing dialog component?How would you evaluate
View Full Document
Unlocking...