A TYPICAL SPEECH RECOGNITION SYSTEM
A typical system decomposes sentences into words, words into phone sequences, and phones into HMM models: