MASc Seminar: A framework for human motion strategy identification and analysisExport this event to calendar

Monday, April 15, 2019 — 10:00 AM EDT

Candidate: Muhammad Umar Choudry

Title: A framework for human motion strategy identification and analysis

 

Date: April 15, 2019

Time: 10:00 AM

Place: EIT 3142

Supervisor(s): Kulic, Dana

 

Abstract:

The human body has many biomechanical degrees of freedom and thus multiple movement strategies can be employed to execute any given task. Automated identification and classification of these movement strategies have potential applications in various fields including sports performance research, rehabilitation, and injury prevention. For example, in the field of rehabilitation, the choice of movement strategy can impact joint loading patterns and risk of injury. The problem of identifying movement strategies is related to the problem of classifying variations in the observed motions. When these variations are macro in nature they are considered to be different movement strategies where the observed movements take on very different trajectories.

Conversely, when the differences between observed movements are more micro in nature they are considered to be part of the same movement strategy where the movements differ only by slight perturbations in the trajectory space. In the simplest scenario, a movement strategy can represent a cluster of similar movement trajectories; but in more complicated scenarios, differences in movements could also lie on a continuum. The goal of this thesis is to develop a computational framework to automatically recognize different movement strategies for performing a task and to identify what makes each strategy different.

 

The proposed framework utilizes Gaussian Process Dynamical Models (GPDM) to convert human motion trajectories from their original high dimensional representation to a trajectory in a lower dimensional space (i.e. the latent space). The dimensionality of the latent space is determined by iteratively increasing the dimensionality until the reduction in reconstruction error between iterations becomes small. Then, the lower dimensional trajectories are clustered using a Hidden Markov Model (HMM) clustering algorithm to identify movement strategies in an unsupervised manner. Next, we introduce an HMM-based technique for detecting differences in observation space variables. This technique is used to detect differences in the latent space variables between the low-dimensional trajectory models and to detect differences in the variables (joint or DoF) between the corresponding high-dimensional (original) trajectory models. Then through correlating latent space variable differences and the DoF differences movement synergies are discovered.

Location 
EIT
Room 3142
200 University Avenue West

Kitchener, ON N2L 3G1
Canada

S M T W T F S
29
30
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
1
2
  1. 2019 (234)
    1. December (1)
    2. November (3)
    3. October (15)
    4. September (26)
    5. August (26)
    6. July (40)
    7. June (24)
    8. May (23)
    9. April (35)
    10. March (25)
    11. February (9)
    12. January (10)
  2. 2018 (150)
    1. December (13)
    2. November (25)
    3. October (12)
    4. September (13)
    5. August (7)
    6. July (23)
    7. June (9)
    8. May (6)
    9. April (9)
    10. March (16)
    11. February (10)
    12. January (7)
  3. 2017 (212)
  4. 2016 (242)
  5. 2015 (242)
  6. 2014 (268)
  7. 2013 (192)
  8. 2012 (31)