東華大學圖書館 |

Language: English

Help

回圖書館首頁

手機版館藏查詢

Back

Switch To: Labeled | MARC Mode | ISBD

Linked to FindBook

Google Book

Amazon

博客來

Objective Evaluation of Young Guide Dogs to Predict Training Success.

Record Type:	Electronic resources : Monograph/item
Title/Author:	Objective Evaluation of Young Guide Dogs to Predict Training Success./
Author:	Mealin, Sean P.
Published:	Ann Arbor : ProQuest Dissertations & Theses, : 2021,
Description:	124 p.
Notes:	Source: Dissertations Abstracts International, Volume: 83-07, Section: B.
Contained By:	Dissertations Abstracts International83-07B.
Subject:	Electrocardiography. -
Online resource:	http://pqdd.sinica.edu.tw/twdaoapp/servlet/advanced?query=28552576
ISBN:	9798522941635

Objective Evaluation of Young Guide Dogs to Predict Training Success.
Mealin, Sean P.

Objective Evaluation of Young Guide Dogs to Predict Training Success. - Ann Arbor : ProQuest Dissertations & Theses, 2021 - 124 p.

Source: Dissertations Abstracts International, Volume: 83-07, Section: B.

Thesis (Ph.D.)--North Carolina State University, 2021.

This item must not be sold to any third party vendors.

Training a guide dog is an expensive and difficult process. Most schools spend approximately $50,000 training each dog, but only successfully graduate 37.5% of dogs that enter the program. At Guiding Eyes for the Blind, a prominent school based in New York state, staff start assessing the suitability of dogs as early as seven-weeks-old. The first of many evaluations, the temperament test has two main goals. First, it allows staff to observe how a dog naturally reacts to novel stimuli and to remove dogs with temperaments incompatible with guide work as early as possible, preserving the resources that would have been expended on their training. Second, it allows them to quantify the temperament traits of promising dogs to select the best candidates to breed the next generation of potential guide dogs. While current evaluation methods are undeniably effective, they are prone to human bias and depend on highly experienced personnel to score the results. Making the scoring of the temperament test more objective would increase the consistency and accuracy of the test results, eliminate errors inherent to subjective interpretation of the test data, and reduce the reliance on domain experts.To achieve the benefits of a more objective temperament test, this dissertation presents both a hardware-based sensor platform and software framework capable of generating the same output as the human scorers. The hardware platform is designed to be worn by the dog undergoing evaluation and captures physiological, inertial, and other synchronized data streams. Since each of the twenty-nine temperament scores are presented as an integer scale from one to five, we developed the framework to treat the prediction task as an ordinalregression problem, which exhibits traits of both a classification and regression task. The framework utilizes supervised deep learning techniques to predict the scores from the time series data collected by the sensors. While we find that batching the temporal data into small windows and using a LSTM network is a valid approach, we show that normalizing the length of the recordings and using a convolutional network produces more accurate results with reduced training times.To prove the efficacy of this solution, we deployed the hardware at Guiding Eyes for the Blind to collect data on approximately seven hundred dogs over a duration of two years. We show that we can use the framework to predict temperament test scores with 92% accuracy using electrocardiography data and 93% accuracy using inertial data. For both experiments, we compare our results to a simple classifier to ensure that our high accuracy is not an artifact of low-variance test outcomes. We also use the framework to explore the feasibility of predicting overall success but find that the size of our current data corpus is insufficient for conclusive results due to the extended data collection time for successful dogs.

ISBN: 9798522941635Subjects--Topical Terms:

773588
Electrocardiography.

Objective Evaluation of Young Guide Dogs to Predict Training Success.
LDR:03961nmm a2200313 4500 001 2342848
005 20220415153022.5
008 241004s2021 ||||||||||||||||| ||eng d
020 $a 9798522941635
035 $a (MiAaPQ)AAI28552576
035 $a (MiAaPQ)NCState_Univ18402038590
035 $a AAI28552576
040 $a MiAaPQ $c MiAaPQ
100 1 $a Mealin, Sean P. $3 3681237
245 1 0 $a Objective Evaluation of Young Guide Dogs to Predict Training Success.
260 1 $a Ann Arbor : $b ProQuest Dissertations & Theses, $c 2021
300 $a 124 p.
500 $a Source: Dissertations Abstracts International, Volume: 83-07, Section: B.
500 $a Advisor: Roberts, David.
502 $a Thesis (Ph.D.)--North Carolina State University, 2021.
506 $a This item must not be sold to any third party vendors.
520 $a Training a guide dog is an expensive and difficult process. Most schools spend approximately $50,000 training each dog, but only successfully graduate 37.5% of dogs that enter the program. At Guiding Eyes for the Blind, a prominent school based in New York state, staff start assessing the suitability of dogs as early as seven-weeks-old. The first of many evaluations, the temperament test has two main goals. First, it allows staff to observe how a dog naturally reacts to novel stimuli and to remove dogs with temperaments incompatible with guide work as early as possible, preserving the resources that would have been expended on their training. Second, it allows them to quantify the temperament traits of promising dogs to select the best candidates to breed the next generation of potential guide dogs. While current evaluation methods are undeniably effective, they are prone to human bias and depend on highly experienced personnel to score the results. Making the scoring of the temperament test more objective would increase the consistency and accuracy of the test results, eliminate errors inherent to subjective interpretation of the test data, and reduce the reliance on domain experts.To achieve the benefits of a more objective temperament test, this dissertation presents both a hardware-based sensor platform and software framework capable of generating the same output as the human scorers. The hardware platform is designed to be worn by the dog undergoing evaluation and captures physiological, inertial, and other synchronized data streams. Since each of the twenty-nine temperament scores are presented as an integer scale from one to five, we developed the framework to treat the prediction task as an ordinalregression problem, which exhibits traits of both a classification and regression task. The framework utilizes supervised deep learning techniques to predict the scores from the time series data collected by the sensors. While we find that batching the temporal data into small windows and using a LSTM network is a valid approach, we show that normalizing the length of the recordings and using a convolutional network produces more accurate results with reduced training times.To prove the efficacy of this solution, we deployed the hardware at Guiding Eyes for the Blind to collect data on approximately seven hundred dogs over a duration of two years. We show that we can use the framework to predict temperament test scores with 92% accuracy using electrocardiography data and 93% accuracy using inertial data. For both experiments, we compare our results to a simple classifier to ensure that our high accuracy is not an artifact of low-variance test outcomes. We also use the framework to explore the feasibility of predicting overall success but find that the size of our current data corpus is insufficient for conclusive results due to the extended data collection time for successful dogs.
590 $a School code: 0155.
650 4 $a Electrocardiography. $3 773588
650 4 $a Computer science. $3 523869
650 4 $a Animal sciences. $3 3174829
690 $a 0984
690 $a 0475
710 2 $a North Carolina State University. $3 1018772
773 0 $t Dissertations Abstracts International $g 83-07B.
790 $a 0155
791 $a Ph.D.
792 $a 2021
793 $a English
856 4 0 $u http://pqdd.sinica.edu.tw/twdaoapp/servlet/advanced?query=28552576