Is simple voice synthesis (just digits 0-9) possible with Cortex-M0+ with 4kB flash?

Question

I'm working on a little side project with the LPC810 (Cortex M0+ 32 bit ARM MCU, 8 pin DIP package, 4kB flash, 1kB SRAM). I need a UI to convey to the user a temperature setting. I only have about 2 pins spare to implement the UI.... so using LEDs is going to a challenge.

I had an idea of using just one pin to drive a speaker or peizzo buzzer and convey the temperature in speech. Just the digits 0 - 9 in english is all I need. Storing 0-9 in PCM is out of the question with so little flash. So I was wondering is anyone aware of open source code that would allow for very high compression of a small dictionary of words what would be light enough on CPU/RAM to work with a Cortex M0 class MCU? (I have a strong hunch this is plain out of reach of low end Cortex M0 MCUs without external memory).

Answer 1

While this may not be a complete, independent and ready-to-use answer, but I think you can get some neat ideas, implement (or port) it on Cortex-M0. Here I assume that from a computation power and resource standpoint, a Cortex-M0 has more to offer, than the popular Atmel 8-bit AVR (ATmega328P) running on an Arduino.

Here are 2 projects, that manage to use the PWM pin of Arduino and an RC-filter circuit to play out synthesized speech. Of course, we are not looking at hi-fidelity audio, but something that is recognizable. Also do note, that apart from the need for a PWM capable pin, your micro-controller might be very busy during the synthesis, so much so that, it might spend most of it's cycles doing it. Software PWM would put further strain.

Now for the 2 projects:

• Cantarino
• Port of Webbot's speech synthesis to Arduino

PS> Personally, I've not implemented them, but looked at them for a project.

Answer 2

I need a UI to convey to the user a temperature setting. I only have about 2 pins spare to implement the UI.... so using LEDs is going to a challenge.

Use the two pins as I²C Bus, and connect a GPIO expansion chip like MAX6956, which has 20 output pins in a 28-pin DIP package. That is enough for a few 7-segment displays, and you can connect more than one expansion chip to the bus.

Answer 3

It may be possible to write code that would fit in 4K that would produce vaguely-recognizable speech for the digits 0-9, but it would probably be easier to define beep patterns for them. A small amount of training might be required to recognize the beep patterns (among other things, if you're using a positional notation, knowing how a zero would be conveyed) but if all temperatures are positive it may not be too hard to define beep sequences for values 90-10 and 9-1.

+90: XXXXX--XX--XX---
+80: XXXXX---X-X-X---
+70: XXXXX-X-X---
+60: XXXXX-X-
+50: XXXXX---
+40: XX--XX--
+30: X-X-X---
+20: X-X-
+10: X---

Each "X" represents 50ms on and 50ms off, while each "-" represents 100ms off. For values 9-1, use the same cadences as 90-10 but a different pitch. The cadence patterns above could easily be stored in an array of 16-bit integers. The sequences above are chosen not just to allow counting, but also to be recognizable as patterns. Counting straight non-rhythmic pulses up to nine can be a bit difficult, but arranging the things to be counted into rhytmic grouping can help a lot.

Answer 4

If board space is not a limitation, A port expansion + few counters (74HC161 or similar), external sample ROM and a DAC can implement this without requiring much code space.