When a muscle contracts it gets fatter with constant volume. If one assumes an ideal scenario of a stationary cylindrical muscle, one could say that the diameter change should have an inverse square relation to muscle length. The voltage of our sensors is also an inverse square relation to distance. Therefore, in an ideal scenario voltage should be linearly proportional with muscle length.
The anatomical position alone controls sensor voltage. At different forces in isometric contraction (same length/anatomical position), the voltage doesn't change. Furthermore, a multivariate linear regression was able to use the combined data of 64 sensors to accurately predict the gesture made within a sampled dataset. Therefore, we believe the simplified description above does an adequate job of explaining why these sensors work.