Programmable synthesis using a new "Speech microprocessor"

In the past, real-time speech synthesis has been achieved using either relatively expensive special hardware or custom integrated circuitry which implements a single synthesis algorithm. In order to fill a need for low cost, high quality synthesis with algorithm flexibility, two NMOS LSI devices which implement a programmable "speech microprocessor" have been developed. By orienting the architecture toward common features of several classes of digital vocal tract models, synthesis computations involving a large number of multiplies, adds, and other operations can be performed efficiently. For example, the processor can implement a complex cascade/parallel formant synthesizer or a lattice filter with up to fourteen poles. The LSI chip set will become commerically available in mid-1980.