Alternatively to pressure sensors, accelerometer-based systems were later developed [39].De Gauna et. al [45]
proposed three methods, with different algorithmic complexities: an improved linear filtering, that 
computes compression depth signal from acceleration signal by using a designed band-pass filter that 
performs double-integration; the detection of the zero-crossing instants in the velocity signal, which 
computes velocity signal from acceleration signal by performing single-integration and calculates 
compression depth and rate from the analysis of zero crossing instants; and the spectral analysis of the 
acceleration signal, which computes compression depth and rate directly from the acceleration signal 
by using the fast Fourier transform.Among the discussed alternatives, the algorithm based on the 
spectral analysis of the acceleration signal provided the most accurate results, since it computes 
compression depth and rate directly from the acceleration signal, avoiding integration errors, although
it has the highest algorithmic complexity.However, since all 
accelerometers are sensible to movement-associated noise, double-integrating the acceleration signal is 
prone to cumulative errors which cause a significant drift in displacement signal, impeding an accurate 
estimation of compression depth [39].Other errors are 
introduced in compression depth calculation by the use of accelerometers, with origin on the compliant 
surface effect [39], and on sensor tilting during compressions, since gravitational acceleration 
component is dependent of sensor orientation [39].Some developments have been made to surpass the 
integration cumulative errors by using signal processing techniques over the extracted acceleration 
signal.