Another thing for you to consider. A number of years ago, I "advised" a group of BME seniors doing a capstone project to create a "paralysis sensor" for use by anesthesiologists during human surgery. There is a test called "Train of Four" to monitor paralysis during surgical procedures -- you deliver a train of 4 small pulses to a peripheral nerve (electrodes on the skin) and count the number of twitches you see in the muscle.
Their idea was to put a small accelerometer on the muscle (taped down) and graphically display the acceleration pulses for the four twitches. These accelerometers delivered X, Y, and Z as voltages, with (I believe) a gain of 0.5 V/g and an offset of 1.5 V. The other interesting fact was that gains and offsets were ±10%. To my chagrin, this rather large and unknown "uncertainty" in the 6 parameters of the accelerometer caused noone on the team to ask "How do we calibrate this device?". They did get that they should make it "orientation-independent" by computing the length of the "acceleration vector", but were not bothered by the fact that when the accelerometer was not moving (flat on a table), you did not get readings of Z = ±1g, X = Y = 0g. Oh, well, at least they could use a USB-6008 ...
Accelerometer chips are fairly inexpensive. You should be able to find an Analog chip which outputs voltages instead of SPI or I²C.
Bob Schor
