All the things you need can be done in software. Maybe a DSP processor is
not a bad idea. But you may even be able to do these things on a
microcontroller like PIC. DTMF dialing on a PIC or a DSP processor is not
difficult. I guess there is already a routine available for generating
sine waves on a PIC. DTMF decoding can also be done by the chip. You don't
need a 8870. There are a lot of algorithms for DTMF detection. Currently I
am working on a DTMF decoding algorithm on a microcontroller. I am working
with Shawn McCaslin of Cicada Semiconductor. Shawn also has a patent on
detecting voice in an environment of noise and voice. For more details
contact him at srm@cicada-semi.com

So I guess you can implement all this on a single chip. The only thing you
need to take care of is the speech digitization.

Amey

-------------------------------------------------------------------------------
                        AMEY A. DEOSTHALI
-------------------------------------------------------------------------------
Lab Address:                              Home Address:
ESPL, ENS526E                             3110 Red River, Apt # 208
Dept. of ECE                              Austin, TX 78705
University of Texas at Austin,            Tel:(512)-499-8957
Austin, TX 78712 - 1084.
Tel:(512)-232-2769
                Email:amey@vision.ece.utexas.edu
                Web  :http://anchovy.ece.utexas.edu/~amey
-------------------------------------------------------------------------------