Functional form of motion priors in human motion perception

Hongjing LU, Tungyou LIN, Lap Fai, Alan LEE, Luminita VESE

Research output: Book Chapters | Papers in Conference ProceedingsConference paper (refereed)Researchpeer-review

2 Citations (Scopus)

Abstract

It has been speculated that the human motion system combines noisy measurements with prior expectations in an optimal, or rational, manner. The basic goal of our work is to discover experimentally which prior distribution is used. More specifically, we seek to infer the functional form of the motion prior from the performance of human subjects on motion estimation tasks. We restricted ourselves to priors which combine three terms for motion slowness, first-order smoothness, and second-order smoothness. We focused on two functional forms for prior distributions: L2-norm and L1-norm regularization corresponding to the Gaussian and Laplace distributions respectively. In our first experimental session we estimate the weights of the three terms for each functional form to maximize the fit to human performance. We then measured human performance for motion tasks and found that we obtained better fit for the L1-norm (Laplace) than for the L2-norm (Gaussian). We note that the L1-norm is also a better fit to the statistics of motion in natural environments. In addition, we found large weights for the second-order smoothness term, indicating the importance of high-order smoothness compared to slowness and lower-order smoothness. To validate our results further, we used the best fit models using the L1-norm to predict human performance in a second session with different experimental setups. Our results showed excellent agreement between human performance and model prediction – ranging from 3% to 8% for five human subjects over ten experimental conditions – and give further support that the human visual system uses an L1-norm (Laplace) prior.
Original languageEnglish
Title of host publicationAdvances in neural information processing systems, 23 : 24th Annual Conference on Neural Information Processing Systems 2010, December 6-9, 2010, Vancouver, B.C., Canada
PublisherNeural Information Processing Systems
Pages1495-1503
Number of pages9
ISBN (Print)9781617823800
Publication statusPublished - 1 Jan 2010
Externally publishedYes

Fingerprint

Motion estimation
Statistics

Bibliographical note

Alan YUILLE, University of California

Cite this

LU, H., LIN, T., LEE, L. F. A., & VESE, L. (2010). Functional form of motion priors in human motion perception. In Advances in neural information processing systems, 23 : 24th Annual Conference on Neural Information Processing Systems 2010, December 6-9, 2010, Vancouver, B.C., Canada (pp. 1495-1503). Neural Information Processing Systems.
LU, Hongjing ; LIN, Tungyou ; LEE, Lap Fai, Alan ; VESE, Luminita. / Functional form of motion priors in human motion perception. Advances in neural information processing systems, 23 : 24th Annual Conference on Neural Information Processing Systems 2010, December 6-9, 2010, Vancouver, B.C., Canada. Neural Information Processing Systems, 2010. pp. 1495-1503
@inproceedings{29ab4c2947364b62aac7891e322b06dc,
title = "Functional form of motion priors in human motion perception",
abstract = "It has been speculated that the human motion system combines noisy measurements with prior expectations in an optimal, or rational, manner. The basic goal of our work is to discover experimentally which prior distribution is used. More specifically, we seek to infer the functional form of the motion prior from the performance of human subjects on motion estimation tasks. We restricted ourselves to priors which combine three terms for motion slowness, first-order smoothness, and second-order smoothness. We focused on two functional forms for prior distributions: L2-norm and L1-norm regularization corresponding to the Gaussian and Laplace distributions respectively. In our first experimental session we estimate the weights of the three terms for each functional form to maximize the fit to human performance. We then measured human performance for motion tasks and found that we obtained better fit for the L1-norm (Laplace) than for the L2-norm (Gaussian). We note that the L1-norm is also a better fit to the statistics of motion in natural environments. In addition, we found large weights for the second-order smoothness term, indicating the importance of high-order smoothness compared to slowness and lower-order smoothness. To validate our results further, we used the best fit models using the L1-norm to predict human performance in a second session with different experimental setups. Our results showed excellent agreement between human performance and model prediction – ranging from 3{\%} to 8{\%} for five human subjects over ten experimental conditions – and give further support that the human visual system uses an L1-norm (Laplace) prior.",
author = "Hongjing LU and Tungyou LIN and LEE, {Lap Fai, Alan} and Luminita VESE",
note = "Alan YUILLE, University of California",
year = "2010",
month = "1",
day = "1",
language = "English",
isbn = "9781617823800",
pages = "1495--1503",
booktitle = "Advances in neural information processing systems, 23 : 24th Annual Conference on Neural Information Processing Systems 2010, December 6-9, 2010, Vancouver, B.C., Canada",
publisher = "Neural Information Processing Systems",

}

LU, H, LIN, T, LEE, LFA & VESE, L 2010, Functional form of motion priors in human motion perception. in Advances in neural information processing systems, 23 : 24th Annual Conference on Neural Information Processing Systems 2010, December 6-9, 2010, Vancouver, B.C., Canada. Neural Information Processing Systems, pp. 1495-1503.

Functional form of motion priors in human motion perception. / LU, Hongjing; LIN, Tungyou; LEE, Lap Fai, Alan; VESE, Luminita.

Advances in neural information processing systems, 23 : 24th Annual Conference on Neural Information Processing Systems 2010, December 6-9, 2010, Vancouver, B.C., Canada. Neural Information Processing Systems, 2010. p. 1495-1503.

Research output: Book Chapters | Papers in Conference ProceedingsConference paper (refereed)Researchpeer-review

TY - GEN

T1 - Functional form of motion priors in human motion perception

AU - LU, Hongjing

AU - LIN, Tungyou

AU - LEE, Lap Fai, Alan

AU - VESE, Luminita

N1 - Alan YUILLE, University of California

PY - 2010/1/1

Y1 - 2010/1/1

N2 - It has been speculated that the human motion system combines noisy measurements with prior expectations in an optimal, or rational, manner. The basic goal of our work is to discover experimentally which prior distribution is used. More specifically, we seek to infer the functional form of the motion prior from the performance of human subjects on motion estimation tasks. We restricted ourselves to priors which combine three terms for motion slowness, first-order smoothness, and second-order smoothness. We focused on two functional forms for prior distributions: L2-norm and L1-norm regularization corresponding to the Gaussian and Laplace distributions respectively. In our first experimental session we estimate the weights of the three terms for each functional form to maximize the fit to human performance. We then measured human performance for motion tasks and found that we obtained better fit for the L1-norm (Laplace) than for the L2-norm (Gaussian). We note that the L1-norm is also a better fit to the statistics of motion in natural environments. In addition, we found large weights for the second-order smoothness term, indicating the importance of high-order smoothness compared to slowness and lower-order smoothness. To validate our results further, we used the best fit models using the L1-norm to predict human performance in a second session with different experimental setups. Our results showed excellent agreement between human performance and model prediction – ranging from 3% to 8% for five human subjects over ten experimental conditions – and give further support that the human visual system uses an L1-norm (Laplace) prior.

AB - It has been speculated that the human motion system combines noisy measurements with prior expectations in an optimal, or rational, manner. The basic goal of our work is to discover experimentally which prior distribution is used. More specifically, we seek to infer the functional form of the motion prior from the performance of human subjects on motion estimation tasks. We restricted ourselves to priors which combine three terms for motion slowness, first-order smoothness, and second-order smoothness. We focused on two functional forms for prior distributions: L2-norm and L1-norm regularization corresponding to the Gaussian and Laplace distributions respectively. In our first experimental session we estimate the weights of the three terms for each functional form to maximize the fit to human performance. We then measured human performance for motion tasks and found that we obtained better fit for the L1-norm (Laplace) than for the L2-norm (Gaussian). We note that the L1-norm is also a better fit to the statistics of motion in natural environments. In addition, we found large weights for the second-order smoothness term, indicating the importance of high-order smoothness compared to slowness and lower-order smoothness. To validate our results further, we used the best fit models using the L1-norm to predict human performance in a second session with different experimental setups. Our results showed excellent agreement between human performance and model prediction – ranging from 3% to 8% for five human subjects over ten experimental conditions – and give further support that the human visual system uses an L1-norm (Laplace) prior.

UR - http://commons.ln.edu.hk/sw_master/4488

M3 - Conference paper (refereed)

SN - 9781617823800

SP - 1495

EP - 1503

BT - Advances in neural information processing systems, 23 : 24th Annual Conference on Neural Information Processing Systems 2010, December 6-9, 2010, Vancouver, B.C., Canada

PB - Neural Information Processing Systems

ER -

LU H, LIN T, LEE LFA, VESE L. Functional form of motion priors in human motion perception. In Advances in neural information processing systems, 23 : 24th Annual Conference on Neural Information Processing Systems 2010, December 6-9, 2010, Vancouver, B.C., Canada. Neural Information Processing Systems. 2010. p. 1495-1503