The microprogramming of pipelined processors
ISCA '77 Proceedings of the 4th annual symposium on Computer architecture
Programming a microcoded processor for speech waveform generation
MICRO 11 Proceedings of the 11th annual workshop on Microprogramming
| Hi-index | 0.00 |
The system for generation of human quality speech from unrestricted English text is under development at MIT. One component of the system is the computationally intense task of converting a parameter stream to a digital speech waveform. This is done using a computational model of the human vocal tract. Largely due to the amount of computation done in this synthesis, the overall system runs much slower than real time. A medium speed, special purpose, digital signal processor has been constructed to simulate the resonant characteristics of the human vocal tract, and so allow the computation to be done in real time. We describe some of the architectural details of the processor, and its usefulness in the task of speech generation.