Patent US 4,331,836

A system incorporating an integrated circuit device or chip which digitally synthesizes human speech using a linear predictive filter. The linear predictive filter comprises a single filter stage only which contains a single multiplier for selectively multiplying a plurality of coefficients, initiating the multiplication of one coefficient at a time, by using a feedback loop with multiplexing techniques so as to input multiplexed signals to the multiplier--as contrasted to a cascade of filter stages. Thus, the single multiplier of the linear predictive filter is utilized repetitively to provide the calculations required. The system also includes a memory for storage of digital filter coefficients, a controller for selectively accessing the coefficients, and a speaker for generating audible sounds.