Voice Command

1) Pick your audio output device. This will almost certainly involve some sort of D/A conversion, as well as appropriate other EE magic to get a speaker of some sort to make noise.

2) Write code to drive your D/A converter from sampled digital sound stored in your microcontroller memory. You'll need to pick a storage format: maybe just straight sampled data, maybe compressed (MP3?).

3) Choose someone whose voice you like (perhaps a professional voice actor, even), and sample their voice while they read all the possible things you want your system to say. You'll probably need to save space by storing just fragments of the output. That is, you'd probably have the sounds "March", "3", and "2003", rather than one big "March 3, 2003", so that you can get by with 12 month samples and 31 day samples rather than 365 samples to name every day of the year.

You'll get to make some quality/cost tradeoffs here. It's better to have your system use more natural phrases like "twelve thirty" instead of "twelve and three zero minutes", but you'll need more samples and a bit smarter code to do more than just read out the digits.

4) Write your application to read your TOD clock, and then output the appropriate fragments of sounds through the D/A.

Details are left as an exercise for the reader.

1) Pick your audio output device. This will almost certainly involve some sort of D/A conversion, as well as appropriate other EE magic to get a speaker of some sort to make noise.

2) Write code to drive your D/A converter from sampled digital sound stored in your microcontroller memory. You'll need to pick a storage format: maybe just straight sampled data, maybe compressed (MP3?).

3) Choose someone whose voice you like (perhaps a professional voice actor, even), and sample their voice while they read all the possible things you want your system to say. You'll probably need to save space by storing just fragments of the output. That is, you'd probably have the sounds "March", "3", and "2003", rather than one big "March 3, 2003", so that you can get by with 12 month samples and 31 day samples rather than 365 samples to name every day of the year.

You'll get to make some quality/cost tradeoffs here. It's better to have your system use more natural phrases like "twelve thirty" instead of "twelve and three zero minutes", but you'll need more samples and a bit smarter code to do more than just read out the digits.

4) Write your application to read your TOD clock, and then output the appropriate fragments of sounds through the D/A.

Details are left as an exercise for the reader.