def on_frame(self, controller): """ Function called on every frame by the leap motion controller Hand position analysis is processed here and vehicle rc channels are then overridden accordingly :param controller: Leap Motion controller :return: None """ hands = controller.frame().hands debug("last hand palm normal: " + str(self.last_hand_palm_normal)) # We only do something if there is one hand detected if len(hands) == 1: palm_normal = hands[0].palm_normal if _debug: debug(str(palm_normal)) channels_override = \ {'1': RC_AVG_VALUE - palm_normal.x * (RC_AVG_VALUE - RC_MIN_VALUE), # RC1: Roll '2': RC_AVG_VALUE - (palm_normal.z + Z_VALUE_CORRECTION) * (RC_AVG_VALUE - RC_MIN_VALUE)} # RC2: Pitch if _rc_throttle: debug("Overriding throttle RC channel to " + str(_rc_throttle)) channels_override[ '3'] = _rc_throttle # RC3: throttle, required to be at the middle to keep altitude # Update the channels override only if there is # a significant change in the hand orientation if (not self.last_hand_palm_normal) \ or not (self.vectors_equal_with_accurary(self.last_hand_palm_normal, palm_normal, HAND_MOVEMENT_DETECTION_THRESHOLD)): debug("Sending drone override RC channels message: " + str(channels_override)) self.vehicle.channels.overrides = channels_override self.last_hand_palm_normal = palm_normal else: # We disable RC 1 and 2 override to stabilize the copter channels_override = \ {'1': None, '2': None} if _rc_throttle: channels_override['3'] = _rc_throttle palm_normal = Vector() palm_normal.x = 0 palm_normal.y = 0 palm_normal.z = 0 # Update only if changed if (not self.last_hand_palm_normal) \ or not (self.vectors_equal_with_accurary(self.last_hand_palm_normal, palm_normal, HAND_MOVEMENT_DETECTION_THRESHOLD)): debug("Sending drone override RC channels message: " + str(channels_override)) self.vehicle.channels.overrides = channels_override self.last_hand_palm_normal = palm_normal
def on_frame(self, controller): """ Function called on every frame by the leap motion controller Hand position analysis is processed here and vehicle rc channels are then overridden accordingly :param controller: Leap Motion controller :return: None """ hands = controller.frame().hands debug("last hand palm normal: " + str(self.last_hand_palm_normal)) # We only do something if there is one hand detected if len(hands) == 1: palm_normal = hands[0].palm_normal if _debug: debug(str(palm_normal)) channels_override = \ {'1': RC_AVG_VALUE - palm_normal.x * (RC_AVG_VALUE - RC_MIN_VALUE), # RC1: Roll '2': RC_AVG_VALUE - (palm_normal.z + Z_VALUE_CORRECTION) * (RC_AVG_VALUE - RC_MIN_VALUE)} # RC2: Pitch if _rc_throttle: debug("Overriding throttle RC channel to " + str(_rc_throttle)) channels_override['3'] = _rc_throttle # RC3: throttle, required to be at the middle to keep altitude # Update the channels override only if there is # a significant change in the hand orientation if (not self.last_hand_palm_normal) \ or not (self.vectors_equal_with_accurary(self.last_hand_palm_normal, palm_normal, HAND_MOVEMENT_DETECTION_THRESHOLD)): debug("Sending drone override RC channels message: " + str(channels_override)) self.vehicle.channels.overrides = channels_override self.last_hand_palm_normal = palm_normal else: # We disable RC 1 and 2 override to stabilize the copter channels_override = \ {'1': None, '2': None} if _rc_throttle: channels_override['3'] = _rc_throttle palm_normal = Vector() palm_normal.x = 0 palm_normal.y = 0 palm_normal.z = 0 # Update only if changed if (not self.last_hand_palm_normal) \ or not (self.vectors_equal_with_accurary(self.last_hand_palm_normal, palm_normal, HAND_MOVEMENT_DETECTION_THRESHOLD)): debug("Sending drone override RC channels message: " + str(channels_override)) self.vehicle.channels.overrides = channels_override self.last_hand_palm_normal = palm_normal
def __init__(self, frames, numframes, *args, **kwargs): norm = 1. / numframes self.ts = frames[0].timestamp / 1000000.0 self.av_numhands = sum(len(fr.hands) for fr in frames) * norm # Calculates the hand's average pitch, roll and yaw self.av_pitch = sum(fr.hands[0].direction.pitch for fr in frames) * norm self.av_roll = sum(fr.hands[0].palm_normal.roll for fr in frames) * norm self.av_yaw = sum(fr.hands[0].direction.yaw for fr in frames) * norm # Calculates the hand's average palm position try: self.av_palm_pos = Vector(0, 0, 0) for fr in frames: self.av_palm_pos += fr.hands[0].palm_position self.av_palm_pos *= norm except: self.av_palm_pos = NaN # Calculates the average number of fingers self.av_fingers = sum(len(fr.hands[0].fingers) for fr in frames) * norm # Calculates the hand's average finger tip position try: self.av_tip_pos = Vector(0, 0, 0) for fr in frames: self.av_tip_pos += sum( (fg.tip_position for fg in fr.hands[0].fingers), Vector()) / len(fr.hands[0].fingers) self.av_tip_pos *= norm except: self.av_tip_pos = NaN # Check whether the hand is horizontal if self.av_numhands > 0.5: self.is_horizontal = 0 self.is_vertical = 0 if self.av_pitch > -10 and self.av_pitch < 20: self.is_horizontal = 1 if self.av_pitch > 55 and self.av_pitch < 70: self.is_vertical = 1 if self.av_pitch < -60 and self.av_pitch > -90: self.is_vertical = -1 # Compute the hand's average palm velocity self.av_palm_vel = sum(fr.hands[0].palm_velocity[1] for fr in frames) * norm # Compute the hand's average fingers speed try: self.av_fingers_speed = 0 for fr in frames: self.av_fingers_speed += sum( fg.tip_velocity[1] for fg in fr.hands[0].fingers) / len(fr.hands[0].fingers) self.av_fingers_speed *= norm except: self.av_fingers_speed = NaN
def update(self, frame, data): # Get the required data hand_state = data['leap_state']['current'] hand_state_prev = data['leap_state']['prev'] elapsed = NaN # (s) d_av_tip_pos = Vector(NaN, NaN, NaN) # (mm) velocity = NaN # (mm/s) try: # Computes some differences between the current and previous frames sets d_av_palm_pos = hand_state.av_palm_pos - hand_state_prev.av_palm_pos d_av_fingers = hand_state.av_fingers - hand_state_prev.av_fingers d_av_tip_pos = hand_state.av_tip_pos - hand_state_prev.av_tip_pos # Computes the elapsed time between the current and previous frames sets elapsed = max(0.000001, hand_state.ts - hand_state_prev.ts) # Computes velocity and acceleration velocity = d_av_palm_pos.magnitude / elapsed except: pass if not math.isnan(velocity) and elapsed <= ( self.max_elapsed / 1000) and velocity <= self.max_vel: if d_av_fingers > 0: self.last_change = hand_state.ts if hand_state.ts - self.last_change < (self.finger_pause / 1000): d_av_palm_pos *= 0 # don't move pointer when fingers changes if hand_state.av_numhands >= 0.5 and hand_state.av_fingers >= 1: # The cursor speed will be inversely proportional to the number of fingers detected by the controller d_av_palm_pos *= 1. / hand_state.av_fingers # Move it! self.move(d_av_palm_pos.x, -d_av_palm_pos.y) elif self.active_fist: # Move it! self.move(d_av_palm_pos.x, -d_av_palm_pos.y) if data['actions']['click']: self.click() if data['actions']['press']: self.press() if data['actions']['release']: self.release() if data['actions']['scroll_up']: self.scroll_up(repeats=data['actions']['scroll_up']) if data['actions']['scroll_down']: self.scroll_down(repeats=data['actions']['scroll_down']) if data['actions']['switch_desk']: self.switch_desktop(data['actions']['switch_desk'])
def update(self,fingers=None): self.touched = 0 if (not fingers) or len(fingers) < 1: return # else... for finger in fingers: center = self.get_sphere_center() radius = self.get_sphere_radius() # first decide whether it's touched if finger.tip_position.distance_to(center) <= radius + 20: self.touched += 1 # then slide if necessary if finger.tip_position.distance_to(center) < radius: rel_to_sphere = finger.tip_position - center flattened = Vector(rel_to_sphere.x,0,rel_to_sphere.z) pitch_radians = flattened.angle_to(rel_to_sphere) radius_at_height = math.cos(pitch_radians)*radius # print "%f -> %f" % (math.degrees(pitch_radians),radius_at_height) z_dist = rel_to_sphere.z x_radius = math.sqrt(radius_at_height**2 - z_dist**2) x_dist = x_radius - math.fabs(rel_to_sphere.x) if finger.tip_position.x >= center.x: x_dist *= -1 self.slide_sphere(x_dist)
class RpsListener(Leap.Listener): logger = None finger_names = ['thumb', 'index', 'middle', 'ring', 'pinky'] bone_names = ['Metacarpal', 'Proximal', 'Intermediate', 'Distal'] state_names = ['STATE_INVALID', 'STATE_START', 'STATE_UPDATE', 'STATE_END'] xyz = ["x", "y", "z"] origin = Vector(0, 0, 0) is_measuring = False rock_state = None rps_state = None def on_init(self, controller): if self.logger != None: self.logger.info("Initialized") def on_connect(self, controller): if self.logger != None: self.logger.info("Connected") def on_disconnect(self, controller): if self.logger != None: # Note: not dispatched when running in a debugger. self.logger.info("Disconnected") def on_exit(self, controller): if self.logger != None: self.logger.info("Exited") def set_logger(self, logger): self.logger = logger def get_hand_data(self, hand): # heuristic way hand_data = [None] * 7 hand_data[0] = hand.id hand_data[1] = hand.confidence idx = 2 hand_direction = hand.direction palm_normal = hand.palm_normal for fng in hand.fingers: angle_hand_fng = hand_direction.angle_to( fng.direction) * 180 / math.pi angle_norm_fng = palm_normal.angle_to( fng.direction) * 180 / math.pi if angle_norm_fng > 90: angle_hand_fng = angle_hand_fng * -1 hand_data[idx + fng.type] = angle_hand_fng return hand_data def start_measure(self): self.is_measuring = True def start_rock_check(self): self.rock_state = RockState(self.logger) return self.rock_state def stop_rock_check(self): self.rock_state = None def start_rps_recognition(self): self.rps_state = RpsState(self.logger) return self.rps_state def stop_rps_recognition(self): self.rps_state = None def stop_measure(self): """ Reset all state info """ self.is_measuring = False self.stop_rock_check() self.stop_rps_recognition() def on_frame(self, controller): if self.is_measuring == False: return frame = controller.frame() if len(frame.hands) == 0: #self.logger.info("No data in this frame") return # use first hand hand = frame.hands[0] # they do not contain valid tracking data and do not correspond to a physical entity if hand.is_valid == False: self.logger.info("Not valid hand") return # store hand data to rock state if self.rock_state != None: self.rock_state.update_hand_data(hand) if self.rps_state != None: self.rps_state.update_hand_data(hand)
from Leap import Vector # MAGIIIIIC NUMBERS ! CENTERED_ORIGIN = Vector(0, 200, 0) MAX_X = 250 MAX_Y = 350 MAX_Z = 150 def rescale_position(old): """ Fit the given vector back into the centered coordinate system. Return a vector with coordinates values in [-1;1] """ # Translate new = old - CENTERED_ORIGIN # Scale new.x /= MAX_X new.y /= MAX_Y new.z /= MAX_Z return new
class RpsListener(Leap.Listener): finger_names = ['thumb', 'index', 'middle', 'ring', 'pinky'] bone_names = ['Metacarpal', 'Proximal', 'Intermediate', 'Distal'] state_names = ['STATE_INVALID', 'STATE_START', 'STATE_UPDATE', 'STATE_END'] xyz = ["x", "y", "z"] origin = Vector(0, 0, 0) logger = None log_filename = "" header_len = 0 is_measuring = False data_dir_path = "" def on_init(self, controller): print("Initialized") def on_connect(self, controller): print("Connected") def on_disconnect(self, controller): # Note: not dispatched when running in a debugger. print("Disconnected") def on_exit(self, controller): print("Exited") def init_logger(self, log_filename): now = datetime.now() now_str = now.strftime("%Y%m%d-%H%M%S%f") data_dir_path = os.path.abspath("data/{}".format(now_str)) os.makedirs(data_dir_path) self.data_dir_path = data_dir_path self.log_filename = log_filename # initialize logger self.logger = self.setup_logger(self.data_dir_path, self.log_filename) data_header = self.create_data_header() data_header_str = "\t".join(data_header) self.header_len = len(data_header) # write header of tsv self.logger.info(data_header_str) def setup_logger(self, data_dir_path, log_filename, level=DEBUG): logger = getLogger(__name__) logger.setLevel(level) # create log in tsv fh = FileHandler( filename="{}/{}.tsv".format(data_dir_path, log_filename)) fm = Formatter("%(asctime)s\t%(message)s") fh.setLevel(DEBUG) fh.setFormatter(fm) logger.addHandler(fh) # don't propagate to parent logger logger.propagate = False return logger def create_data_header(self): data_header = [] # hand_header = 7 + 3*3 = 16 hand_header = [ "hand_type", "hand_id", "hand_confidence", "palm_x", "palm_y", "palm_z", "palm_width" ] for bs in ["x_basis", "y_basis", "z_basis"]: hand_header.extend( ["hand_{}_{}".format(bs, ax) for ax in self.xyz]) # finger_header = 5 * (2 + 3 + 3 + 4*3) = 100 finger_header = [] for f in self.finger_names: # length and width are decided when your hand is first detected (static) finger_header.extend( ["{}_{}".format(f, t) for t in ["len", "width"]]) finger_header.extend( ["{}_direction_{}".format(f, ax) for ax in self.xyz]) finger_header.extend( ["{}_velocity_{}".format(f, ax) for ax in self.xyz]) # finger bones, all fingers contain 4 bones for b in range(0, 4): finger_header.extend( ["{}_{}_center_{}".format(f, b, ax) for ax in self.xyz]) data_header.extend(hand_header) data_header.extend(finger_header) return data_header def get_hand_data(self, hand): # fast array initializaion d = [None] * self.header_len d[0] = "L" if hand.is_left else "R" d[1] = hand.id d[2] = hand.confidence pp = hand.palm_position d[3] = pp.x d[4] = pp.y d[5] = pp.z d[6] = hand.palm_width idx = 7 basis = hand.basis for bs in [basis.x_basis, basis.y_basis, basis.z_basis]: for val in [bs.x, bs.y, bs.z]: d[idx] = val idx += 1 for fng in hand.fingers: d[idx] = fng.length idx += 1 d[idx] = fng.width idx += 1 fng_dir = fng.direction fng_vel = fng.tip_velocity for val in [fng_dir.x, fng_dir.y, fng_dir.z]: d[idx] = val idx += 1 for val in [fng_vel.x, fng_vel.y, fng_vel.z]: d[idx] = val idx += 1 for b in range(0, 4): bn = fng.bone(b) d[idx] = bn.center.x idx += 1 d[idx] = bn.center.y idx += 1 d[idx] = bn.center.z idx += 1 return d def start_measure(self): self.is_measuring = True def stop_measure(self): self.is_measuring = False def on_frame(self, controller): if self.is_measuring == False: return frame = controller.frame() if len(frame.hands) == 0: print("No data in this frame") return # use first hand hand = frame.hands[0] # they do not contain valid tracking data and do not correspond to a physical entity if hand.is_valid == False: print("Not valid hand") return hand_data = self.get_hand_data(hand) self.logger.info("\t".join(map(str, hand_data)))