Hi all,

We are in a project designing a wrist PDA to be used by elderly people, who
may be
vision impaired.  The user interface has speech feedback.  We plan to use
pre-recorded
digital speech, so that messages can be output in a familiar voice.  But
we'd also like
users or helpers to be able to record their own messages for subsequent
output.

In the future we might use a speech synthesiser - but it will have to be
multilingual,
with German, Italian, and Spanish as well as (British) English.

We urgently need a suitable chip or chip set for handling speech.  The wrist
unit contains
a 66MHz ARM7.  We need to be able to hold at least 6 minutes of speech.  We
envisage
three possible types of solution:

1.  The speech chip contains digital memory for storing one or more
messages.
For output, a message is loaded into the speech chip memory from the ARM's
memory,
and then it is output via DAC to the speaker.  For input, a spoken message
is input
via ADC and recorded in speech memory, from which it is transferred into
the ARM's memory.

2.  The ARM7 has an 8 kHz interrupt, and on interrupt it outputs a speech
sample to
DAC, or inputs a speech sample from ADC.

3.  Intermediate between 1 and 2, the speech chip has, for output via DAC, a
FIFO
which is loaded from ARM memory with speech samples every 10 msec or so.
For input via ADC, the chip has a FIFO which is emptied into ARM memory
every 10 msec or so.

Ideally there should be low pass filters for input and output integrated on
the chip.
And it should be possible to put the chip or chips into an idle mode to save
power when
not in use.

Does anybody on the list have experience with any of these solutions?  Can
anybody
suggest a suitable chip or chipset?  Cost and space are important factors.

Cheers from Chiswick,

John