What did you have in mind? I enjoy building things I'm just a little limited on budget
Ok, simplistically described only because I'm short on time these days, in my working life I spent a lot of time with creating magnetic fields and detecting changes in those fields in a very rough environment; the drilling industry.
The method I'm referring to is not unique and is similar to the method used in a guitar style pickup with a key exception.
One needs a steady (or static) DC magnetic field which can be produced by the correct magnet. Then one needs a way to detect changes to that field which can be done with coiled-wire sensor(s). Then one needs ferrous material to move through that field which is your steel strings. This essentially describes the sensor system, and I'm thinking of a way to do this with a small magnet and an inductor. The shapes and sizes of these components would have to be determined by engineering specs and experimentation, because there are many variables. I'm imagining one sensor under each string (or pair of strings) built using standard electronic components.
The key difference between what you are doing and a generally available guitar pickup is that you need a "digital" sensor, and a guitar pickup is built for "hi-fidelity" and a relatively fixed distance between strings. Another difference is that your hammered strings begin vibration at 90 degrees to a traditional "plucked" string which may have something to do with the timing of the strike. Also, in your case, you have many more strings to detect and therefore the need for many more sensors which suggests some sort of combining of the signals created in the sensors; perhaps some sort of digital ORRING process.
I'm imagining the amplification of the signal from the sensors to be a significant engineering project, but certainly not insurmountable. Once digitized, this combination of individual string strikes could easily be merged into a single output, but significantly you would have the ability to experiment with detecting certain strings more often than other strings.
So, the reason I say "it would require considerable development" is because sensor development is a cyclic process of controlling variables, changing one variable and testing for results until acceptable results are obtained.
That's it in a nutshell. Nothing wild. Just good old solid engineering with some PICAXE thrown in there for fun.