well-known property in signal processing
Sat Jan 6 00:03:51 2018 - permalink -
which states that: if the signal contains a single dominant fre-
quency, then that frequency can be accurately measured by
performing a linear regression on the phase of the complex
time-domain signal . Hence, we perform an additional
optimization step, whereby we filter the output of the FFT,
keeping only the peak and its two adjacent bins; this filtering
allows us to eliminate noise caused by extraneous and non-
periodic movements. Then, we perform an inverse FFT to
obtain a complex time-domain signal
. The phase of
will be linear and its slope will correspond to the breathing
frequency, i.e., the breathing rate.
a human is not a point reflector, our experiments show that
a separation of 1–2 m is needed for high accuracy.