import re
import numpy as np
import pandas
import sys
from config.Skeleton import Skeleton
from config.AnybodyFirstFrame import AnybodyFirstFrame
from config.BasisFirstFrame import BasisFirstFrame
from resources.pymo.pymo.data import MocapData
from RotationUtil import rot2eul, get_order
from resources.LeapSDK.v53_python39 import Leap
class LeapData:
"""
A class to convert LeapMotion frames to PyMO data structure (MocapData)
Calculates translations (offsets) and rotation data for the joints
"""
def __init__(self, channel_setting='rotation', frame_rate=0.033333, anybody_basis=True):
self._skeleton = {}
self._setting = Skeleton(channel_setting)
self._motion_channels = []
self._motions = []
self._root_name = ''
self.data = MocapData()
self.first_frame = None
self.anybody_first_frame = AnybodyFirstFrame()
self.basis_first_frame = BasisFirstFrame()
# anybody_reference = True -> Use Anybody basis from config/*.json (see AnybodyFirstFrame)
# anybody_reference = False -> Use Leap Motion First Frame for basis
self.anybody_basis = anybody_basis
self.status = 0
self._frame_rate = frame_rate
self._skeleton = self._setting.skeleton
# fill channels into skeleton in selected order (i.e. xyz)
self._skeleton_apply_channels(self._setting.channel_setting)
self._root_name = self._setting.root_name
# initialize offsets for each joint
for joint_name, joint_value in self._skeleton.items():
joint_value['offsets'] = [0, 0, 0]
for channel in joint_value['channels']:
self._motion_channels.append((joint_name, channel))
def parse(self):
self.data.skeleton = self._skeleton
self.data.channel_names = self._motion_channels
if not self.first_frame:
sys.exit("No data was recorded - will terminate now!")
self.data.values = self._motion2dataframe()
self.data.root_name = self._root_name
self.data.framerate = self._frame_rate
return self.data
def _check_frame(self, frame):
"""
Check whether frame and hand and fingers are valid, produce error when left hand is shown
"""
status_no_hand = 1
status_no_finger = 2
status_left_hand = 3
status_valid = 4
if frame.hands.is_empty:
if self.status != status_no_hand:
print("-- No hand found. --")
self.status = status_no_hand
return False
# Get the first hand
hand = frame.hands[0]
if hand.is_left:
if self.status != status_left_hand:
print("-- Please use your right hand. --")
self.status = status_left_hand
return False
if not hand.is_right and not hand.is_valid:
return False
# Check if the hand has any fingers
fingers = hand.fingers
if fingers.is_empty:
if self.status != status_no_finger:
print("-- No valid fingers found. --")
self.status = status_no_finger
return False
# frame_number = 0 if not self.first_frame else frame.id - self.first_frame.id
if self.status != status_valid:
print("-- Valid right hand found, recording data. --")
self.status = status_valid
return True
def add_frame(self, frame):
if not self._check_frame(frame):
return None
# Get the first hand
hand = frame.hands[0]
if not self.first_frame:
self.first_frame = frame
channel_values = self._get_channel_values(hand, firstframe=True)
self._motions.append((0, channel_values))
return
channel_values = self._get_channel_values(hand)
self._motions.append((frame.timestamp - self.first_frame.timestamp, channel_values))
return frame
def _get_channel_values(self, hand, firstframe=False):
channel_values = []
# export_basis = {}
for joint_name, joint_value in self._skeleton.items():
# motion data with rotations
if joint_name == self._root_name:
x_pos, y_pos, z_pos = LeapData._get_root_offset()
elif joint_name == 'RightElbow':
x_pos, y_pos, z_pos = LeapData._get_elbow_offset(hand)
elif joint_name == 'RightHand':
x_pos, y_pos, z_pos = LeapData._get_wrist_offset(hand)
else:
x_pos, y_pos, z_pos = LeapData._get_finger_offset(joint_name, hand)
x_rot, y_rot, z_rot = self._calculate_euler_angles(hand, joint_name)
x_rot *= Leap.RAD_TO_DEG
y_rot *= Leap.RAD_TO_DEG
z_rot *= Leap.RAD_TO_DEG
# if joint_name == 'RightHand':
# print(x_rot, y_rot, z_rot)
if firstframe:
if self.anybody_basis:
x_pos, y_pos, z_pos = self._calculate_offset(joint_name, [x_pos, y_pos, z_pos])
joint_value['offsets'] = [x_pos, y_pos, z_pos]
# # dump the basis of leap motion bones
# if 'End' not in joint_name and 'Root' not in joint_name:
# export_basis[joint_name] = np.ndarray.tolist(self._get_basis(hand, joint_name))
for channel in joint_value['channels']:
if channel == 'Xposition':
channel_values.append((joint_name, channel, x_pos))
if channel == 'Yposition':
channel_values.append((joint_name, channel, y_pos))
if channel == 'Zposition':
channel_values.append((joint_name, channel, z_pos))
if channel == 'Xrotation':
channel_values.append((joint_name, channel, x_rot))
if channel == 'Yrotation':
channel_values.append((joint_name, channel, y_rot))
if channel == 'Zrotation':
channel_values.append((joint_name, channel, z_rot))
# # dump the basis of leap motion bones
# if firstframe:
# import json
# import datetime
# with open('../output/{}basis.json'.format(datetime.datetime.today().strftime('%Y%m%d_%H%M%S')), 'w') as o:
# json.dump(export_basis, o)
return channel_values
def _calculate_euler_angles(self, hand, joint_name):
initial_hand = self.first_frame.hands[0]
# special case for root and finger tip
if joint_name == self._root_name or not self._skeleton[joint_name]['children']:
return 0.0, 0.0, 0.0
if self.anybody_basis:
# compare basis to anybody basis
parent_initial_basis = self._get_basis_first_frame(self._skeleton[joint_name]['parent'])
initial_basis = self._get_basis_first_frame(joint_name)
else:
# compare basis to first frame from Leap Motion
parent_initial_basis = self._get_basis(initial_hand, self._skeleton[joint_name]['parent'])
initial_basis = self._get_basis(initial_hand, joint_name)
parent_basis = self._get_basis(hand, self._skeleton[joint_name]['parent'])
basis = self._get_basis(hand, joint_name)
# if joint_name == 'RightHand':
# print(basis)
# calculation of local rotation matrix - important!!!
rot = np.matmul(
np.matmul(
initial_basis, np.transpose(basis)
),
np.transpose(
np.matmul(
parent_initial_basis, np.transpose(parent_basis)
)
)
)
return rot2eul(rot)
def _get_basis(self, hand, joint_name):
if joint_name == self._root_name:
return np.array([[1, 0, 0],
[0, 1, 0],
[0, 0, 1]])
if joint_name == 'RightElbow':
return LeapData._basismatrix(hand.arm.basis)
if joint_name == 'RightHand':
return LeapData._basismatrix(hand.basis)
# else, return basis of the finger
finger, bone_number = LeapData._split_key(joint_name)
fingerlist = hand.fingers.finger_type(LeapData._get_finger_type(finger))
bone = fingerlist[0].bone(LeapData._get_bone_type(bone_number))
return LeapData._basismatrix(bone.basis)
def _get_basis_first_frame(self, joint_name):
if joint_name == self._root_name:
return np.array([[1, 0, 0],
[0, 1, 0],
[0, 0, 1]])
return self.anybody_first_frame.get_basis(joint_name)
def _calculate_offset(self, joint_name, leap_offset):
if joint_name == self._root_name:
return 0, 0, 0
leap_length = np.linalg.norm(leap_offset)
position = self._get_anybody_position(joint_name)
position_parent = self._get_anybody_position(self._skeleton[joint_name]['parent'])
offset_anybody = position - position_parent
# make a unit vector
offset_anybody = np.divide(offset_anybody, np.linalg.norm(offset_anybody))
offset = offset_anybody * leap_length
return offset[0], offset[1], offset[2]
def _get_anybody_position(self, joint_name):
if joint_name == self._root_name:
return np.array([0, 0, 0])
return self.anybody_first_frame.get_position(joint_name)
@staticmethod
def _get_root_offset():
return 0, 0, 0
@staticmethod
def _get_elbow_offset(hand):
arm = hand.arm
return arm.elbow_position.x, arm.elbow_position.y, arm.elbow_position.z
@staticmethod
def _get_wrist_offset(hand):
arm = hand.arm
x_wrist = hand.wrist_position.x - arm.elbow_position.x
y_wrist = hand.wrist_position.y - arm.elbow_position.y
z_wrist = hand.wrist_position.z - arm.elbow_position.z
return x_wrist, y_wrist, z_wrist
@staticmethod
def _get_finger_offset(key, hand):
key, bone_number = LeapData._split_key(key)
fingerlist = hand.fingers.finger_type(LeapData._get_finger_type(key))
# vector between wrist and joint metacarpal proximal (length of carpals)
if bone_number == 1 or ('Thumb' in key and bone_number == 2):
bone = fingerlist[0].bone(LeapData._get_bone_type(bone_number))
return \
bone.prev_joint.x - hand.wrist_position.x, \
bone.prev_joint.y - hand.wrist_position.y, \
bone.prev_joint.z - hand.wrist_position.z
# vector for bones metacarpal, proximal, intermediate, distal
bone = fingerlist[0].bone(LeapData._get_bone_type(bone_number - 1))
return \
bone.next_joint.x - bone.prev_joint.x, \
bone.next_joint.y - bone.prev_joint.y, \
bone.next_joint.z - bone.prev_joint.z
@staticmethod
def _split_key(key):
key_split = re.split('(\d)', key)
key = key_split[0]
if key_split[-1] == '_Nub':
return key, 5
else:
return key, int(key_split[1])
@staticmethod
def _get_finger_type(key):
if key == 'RightHandThumb':
return Leap.Finger.TYPE_THUMB
if key == 'RightHandIndex':
return Leap.Finger.TYPE_INDEX
if key == 'RightHandMiddle':
return Leap.Finger.TYPE_MIDDLE
if key == 'RightHandRing':
return Leap.Finger.TYPE_RING
if key == 'RightHandPinky':
return Leap.Finger.TYPE_PINKY
else:
raise Exception('Key ({}) did not match'.format(key))
@staticmethod
def _get_bone_type(bone_number):
if bone_number == 4:
return Leap.Bone.TYPE_DISTAL
if bone_number == 3:
return Leap.Bone.TYPE_INTERMEDIATE
if bone_number == 2:
return Leap.Bone.TYPE_PROXIMAL
if bone_number == 1:
return Leap.Bone.TYPE_METACARPAL
else:
raise Exception('bone number ({}) did not match'.format(bone_number))
@staticmethod
def _basismatrix(basis):
return np.array([[basis.x_basis.x, basis.y_basis.x, basis.z_basis.x],
[basis.x_basis.y, basis.y_basis.y, basis.z_basis.y],
[basis.x_basis.z, basis.y_basis.z, basis.z_basis.z]])
def _get_channels(self, joint_name, channel_setting):
if '_Nub' in joint_name:
return []
channels_position = ['Xposition', 'Yposition', 'Zposition']
channels_rotation = ['Xrotation', 'Yrotation', 'Zrotation']
order = get_order() # rotation order, i.e. xyz
channels_rotation = \
[channels_rotation[order[0]]] + [channels_rotation[order[1]]] + [channels_rotation[order[2]]]
if channel_setting == 'position' or joint_name == self._root_name:
return channels_position + channels_rotation
return channels_rotation
def _skeleton_apply_channels(self, channel_setting):
for joint_name, joint_dict in self._skeleton.items():
joint_dict['channels'] = self._get_channels(joint_name, channel_setting)
def _motion2dataframe(self):
"""Returns all of the channels parsed from the LeapMotion sensor as a pandas DataFrame"""
time_index = pandas.to_timedelta([f[0] for f in self._motions], unit='s')
frames = [f[1] for f in self._motions]
channels = np.asarray([[channel[2] for channel in frame] for frame in frames])
column_names = ['%s_%s' % (c[0], c[1]) for c in self._motion_channels]
return pandas.DataFrame(data=channels, index=time_index, columns=column_names)