ped_online.r2py

"""
<Program Name>
  pedometer.r2py
<Purpose>
  This is a script for walking step counter. Analysis of the sensor data 
  from accelerometer to detect the walking / running steps. Introducing
  pre-calibration stage, noise level threshold and moving average filter
  to accurate step detection for difference devices.
  *Note: the device must be stable for 1 second pre-calibration from beginning
"""




dy_import_module_symbols('getsensor.r2py')
dy_import_module_symbols('pre_online.r2py')
hanning = dy_import_module('moving_average_hanning.r2py')
butterworth = dy_import_module('butterworth.r2py')


# constants for pre-calibratino stage
PRE_LEARN_SAMPLE_NUMBER = 100
ZERO_INTERVAL = 0.2                  # Minimum time difference between each step in senconds
PEAK_INTERVAL = 0.4




# Peodmeter class, includes filtering raw data and step estimation
class Pedometer:
  def __init__(self, precalibration, height):
    # Coefficient from pre-calibration stage
    self.gravity_constant = precalibration.get_gravity()
    self.samplerate = precalibration.get_sample_rate()
    self.threshold = precalibration.get_threshold()
    if self.threshold < 0.05:
      self.threshold = 0.05
    self.steptime = precalibration.time

    # Initialize moving average filter

 
    self.maf = hanning.HanningFilter(precalibration.time, 0.125)


    # Initialize low pass filter
    self.lpf = butterworth.ButterworthFilter("low", 5.0, self.samplerate, precalibration.time, 0.125)

    self.hpf = butterworth.ButterworthFilter("high", 0.1, self.samplerate, precalibration.time, 0.125)


    self.height = height

    self.zerocount = 0
    self.lastmag = 0.0
    self.peakcount = 0
    self.peak_window_start = self.steptime
    self.maglist = []
    self.timelist = []
    self.max_mag = self.threshold
    self.max_time = self.peak_window_start
    self.last_peak_time = self.max_time
    self.flag = 1

    self.distance_zero = []
    self.time_zero = []
    self.distance_peak = []
    self.time_peak = []

    self.acc_for_method = []
    self.carrying_flag = 0

    self.motion_flag = 0

    self.moving_freq = 0.0

    log("\nThank you for your patiences. Please walk around.\n\nStep detections:\n")

    

    

  def detect_step(self, raw_acc, time, method = 0): 


    # Non-gravity acceleration
    raw_mag = matrix_row_magnitude([raw_acc['xforce'], raw_acc['yforce'], raw_acc['zforce']]) - self.gravity_constant



    # linear phase low pass filter
    lpf_mag = self.lpf.filter(raw_mag, time, True)
    hpf_mag = self.hpf.filter(lpf_mag, time, True)



    # moving average filter
    maf_mag = self.maf.hanning_filter(hpf_mag, time)


    
    # step detection
    peak = self._detect_peak(maf_mag, time)
    zero = self._detect_zero(maf_mag, time)


    # carrying method detection
    self._method_detection(raw_acc)


    self._motion_detect(maf_mag, time)
      

    if method == 1:
      return peak
    else:
      return zero



  def _method_detection(self, acc):
    self.acc_for_method.append(acc)
    if len(self.acc_for_method) == 500:
      avg_acc_x = 0.0
      avg_acc_y = 0.0
      avg_acc_z = 0.0
      for i in range(0, 500):
        avg_acc_x += self.acc_for_method[i]['xforce']
        avg_acc_y += self.acc_for_method[i]['yforce']
        avg_acc_z += self.acc_for_method[i]['zforce']
      avg_acc_x /= 500.0
      avg_acc_y /= 500.0
      avg_acc_z /= 500.0

      # hold in hand
      if avg_acc_z ** 2 > (avg_acc_x ** 2 + avg_acc_y ** 2) * 2:
        self.carrying_flag = 1
      else:
        # trousers
        if avg_acc_y ** 2 > (avg_acc_x ** 2 + avg_acc_z ** 2) * 2:
          self.carrying_flag = 2
        # coat
        else:
          self.carrying_flag = 3
      self.acc_for_method = []

      if self.carrying_flag == 0:
        log("Current carrying method: unknown\n")
      elif self.carrying_flag == 1:
        log("Current carrying method: hold in hand\n")
      elif self.carrying_flag == 2:
        log("Current carrying method: put in trousers pocket\n")
      elif self.carrying_flag == 3:
        log("Current carrying method: put in coat pocket\n")


  def _motion_detect(self, mag, time):
    if mag > self.threshold or time - self.steptime < 1.0:
      if self.motion_flag == 0:
        if self.moving_freq > 2.5:
          log("Status: running\n")
        else:
          log("StatusL walking\n")
      self.motion_flag = 1
    else:
      if self.motion_flag == 1:
        log("Status: standing\n")
      self.motion_flag = 0


  def get_stepcount(self):
    return {"zero": self.zerocount, "peak": self.peakcount} 





  def get_distance(self):
    sum_zero = 0.0
    sum_peak = 0.0
    if len(self.distance_zero) > 0:
      for i in range(0, len(self.distance_zero)):
        sum_zero += self.distance_zero[i]
      
    if len(self.distance_peak) > 0:
      for i in range(0, len(self.distance_peak)):
        sum_peak += self.distance_peak[i]

    # Average of people's stride is height * 0.415
    # Average stride works as reference
    average_zero_distance = self.zerocount * 0.415 * self.height/100
    average_peak_distance = self.peakcount * 0.415 * self.height/100
    return {"sum_zero": sum_zero, "sum_peak": sum_peak, "average_zero_distance": average_zero_distance, "average_peak_distance": average_peak_distance}




  # crossing noise level and actual step interval > minimum step interval
  def _detect_zero(self, data, time):
    stepped = False
    if self.lastmag <= self.threshold and data > self.threshold and time - self.steptime >= ZERO_INTERVAL:
      self.zerocount += 1
      self.distance_zero.append(self._distance_estimation(self.steptime, time))
      self.time_zero.append(time)
      self.steptime = time
      log("By zero crossing algorithm: time:", time, "Current steps count:", self.zerocount, '\n')
      stepped = True
    self.lastmag = data

    return stepped

  


  # a moving window to capture the peak
  # will change to queue later
  def _detect_peak(self, data, time): 
    stepped = False
    self.maglist.append(data)
    self.timelist.append(time)

    if time - self.peak_window_start >= PEAK_INTERVAL * self.flag:
      for i in range(1, len(self.maglist)-1):
        if self.maglist[i-1] <= self.maglist[i] and self.maglist[i] > self.maglist[i+1] and \
            self.maglist[i] > self.max_mag:
          self.max_mag = self.maglist[i]
          self.max_time = self.timelist[i]
          self.flag = 0

      if self.flag:
        self.flag += 1
      else:
        self.flag = 1
        self.peakcount += 1
        stepped = True

        self.distance_peak.append(self._distance_estimation(self.last_peak_time, self.max_time))
        self.time_peak.append(time)
        log("By peak search algorithm: time:", time, "Current steps count:", self.peakcount, '\n')
        self.peak_window_start = time
        self.last_peak_time = self.max_time
        self.max_mag = self.threshold
  # self.last_maxtime = self.maxtime
        self.maglist = []
        self.timelist = []
  
    return stepped





  # Distance estimation by height and steps frequency  
  def _distance_estimation(self, lasttime, currenttime):
    interval = currenttime - lasttime
    height = self.height/100.0
    step_per_second = 1.0/interval
 
    step_length = height * 0.43 + 0.2 * step_per_second - 0.4

    if interval > 1.0 or step_length > 300.0:
      step_length = height * 0.43
    elif interval < 0.25:
      step_length = 0.0


    return step_length

# -*- mode: python;-*-