chattering chips

chattering chips elektor september 1981 — 9-43 ICs that talk There is a simple way to produce synthetic speech. Feed the desired words through an analog-to-digital con- verter and store the output in a memory. As required, this data can be recovered and passed through a digital-to-analog converter to produce a spoken message. Easy, yes, but no-one does it this way. With good reason: it would require at least 64 Kbits of memory for one second of speech! For a viable system, some kind of drastic data reduction is required. Up to this point, all manufacturers are in full agreement; from here on, they all differ. Broadly speaking, there are two main approaches. The first is to make full use of the ex- perience gained in telecommunication systems. Post Office engineers dis- In real life, the techniques used are rather more complicated and less easy to explain. "Signal coding", "Waveshape coding", "Adaptive delta modulation" - you name it, they use it. Quite sen- sational results can ...