CallRecorder -- Toy project to learn about Android programming.

Thread about VOICE_CALL not getting both sides of conversation:
http://code.google.com/p/android/issues/detail?id=2117

Prior to API v4 the possible audio source for recording audio on the
handset was limited to the microphone; in order to do call recording
people resorted to putting the phone in speakerphone mode and just
accepting the crappy recording quality.

API v4 introduced 3 new sources:

VOICE_UPLINK
VOICE_DOWNLINK
VOICE_CALL

which are supposed to provide what you might expect from the names:
what you're saying, what the person on the other end is saying, or
both together respectively.

From what I can tell VOICE_CALL didn't actually work right in v4
(various complaints in the referenced thread and other conversations
online, and the emulator barfs when trying to record from VOICE_CALL
for v4 image but not with later versions.)

At this point I need to play with some real hardware at different SDK
levels to get some real data on that issue.  I had the bright idea to
try recording from UPLINK and DOWNLINK simultaneously on different
threads and mix in software; but I don't know if that will work (or be
necessary) without being able to test on real handsets.

Other things I would like to try out with this testbed:

* Speex encoding by using NDK to build libspeex for ARM and use the
  AudioRecord class rather than MediaRecorder to get raw PCM data to
  feed it.  (Need libogg too?)

* Is it possible to selectively replace the call log part of the built
  in phone stuff with our own to present recording info in the normal
  call log?

* Do encryption (bouncy castle?  openssh/openssl via NDK?) with public
  key protocols so that people can't listen to the call records just
  by getting the handset if you choose not to keep the private key on
  it.  Also, you could encrypt where you don't have the private key at
  all (using your lawyers key?)

* Either use the built in voice recognition or something server side
  to do transcripts/searchability.  Make sure to use location data in
  search stuff as well (ie. search all conversations I had at the
  Library in June kind of thing.)  If it works reasonably well add more
  focus on voice memo and general recording from the mic.

* Use location data to auto-warn people when they are recording in a
  (US) state which requires both sides of the conversation to consent
  to recording.

* More useful recording player component (slider bar to scrub around
  the conversation, make notes associated with the call/particular
  time in the call, etc.)

* Make the UI/configs not suck (default app activity showing icon menu
  + call log, config to exclude/include specific phone numbers,
  reasonable file naming, etc..