fix: remove high-pass prefilter (ssf is already a high-pass)

pasada-lib/step_detector.cpp

@@ -4,24 +4,8 @@
 
 #include "step_detector.h"
 
-// TODO: we are hardcoding filter coefficients for 60 Hz
-// TODO: this is tolerable for 50 Hz
-
-// TODO: check if we can do with floats instead of doubles
-// (check how much the [already bad] accuracy of filtering suffers)
-
-// TODO: in Java, check if delta timestamps effectively match FPS
-// TODO: FPS constant should be passed as argument to C++ (but we keep an FPS define to validate the coefficients)
-
-// Butterworth filter: order=5, fc=0.5, fs=60, btype='highpass'
-static std::vector<double> hpf_taps_b {0.91875845, -4.59379227,  9.18758454, -9.18758454,  4.59379227, -0.91875845};
-static std::vector<double> hpf_taps_a {1.        , -4.83056552,  9.33652742, -9.02545247,  4.36360803, -0.8441171};
-static size_t upslope_width = 4;
-const size_t len_refr = (size_t) (FPS / (MAX_BPM / 60));
-
 StepDetector::StepDetector(double fps, StepListener *listener, bool debug) :
     listener(listener),
-    f_highpass(hpf_taps_b, hpf_taps_a),
     f_neg(1, 0, 0, std::vector<double> {-1.0}),
     f_ssf(fps),
     f_ssd(fps),
@@ -35,8 +19,7 @@ StepDetector::StepDetector(double fps, StepListener *listener, bool debug) :
 
 void StepDetector::filter(std::vector<float> values) {
     // TODO: later on, we should use a vector projection towards gravity
-    auto s0 = (double) values[1]; // take y-axis value for now
-    auto s1 = f_highpass.filter(s0);
+    auto s1 = (double) values[1]; // take y-axis value for now
     auto s2 = f_neg.filter(s1);
     auto s3 = f_ssf.filter(s2);
     auto s4 = f_ssd.filter(s3);