FPGA Implementation of a Pipelined Gaussian Calculation for HMM-Based Large Vocabulary Speech Recognition

Figure 1

Language, lexicon, and phoneme sources combined.