Multi-source domain adaptation and its application to early detection of fatigue

Abstract

We consider the characterization of muscle fatigue through noninvasive sensing mechanism such as surface electromyography (SEMG). While changes in the properties of SEMG signals with respect to muscle fatigue have been reported in the literature, the large variation in these signals across different individuals makes the task of modeling and classification of SEMG signals challenging. Indeed, the variation in SEMG parameters from subject to subject creates differences in the data distribution. In this paper, we propose a transfer learning framework based on the multi-source domain adaptation methodology for detecting different stages of fatigue using SEMG signals, that addresses the distribution differences. In the proposed framework, the SEMG data of a subject represent a domain; data from multiple subjects in the training set form the multiple source domains and the test subject data form the target domain. SEMG signals are predominantly different in conditional probability distribution across subjects. The key feature of the proposed framework is a novel weighting scheme that addresses the conditional probability distribution differences across multiple domains (subjects). We have validated the proposed framework on Surface Electromyogram signals collected from 8 people during a fatigue-causing repetitive gripping activity. Comprehensive experiments on the SEMG data set demonstrate that the proposed method improves the classification accuracy by 20% to 30% over the cases without any domain adaptation method and by 13% to 30% over the existing state-of-the-art domain adaptation methods.

A survey of transfer learning

“…Another area of literature inconsistencies is in characterizing the transfer learning process with respect to the availability of labeled and unlabeled data. For example, Daumé [22] and Chattopadhyay [14] define supervised transfer learning as the case of having abundant labeled source data and limited labeled target data, and semi-supervised transfer learning as the case of abundant labeled source data and no labeled target data. In Gong [42] and Blitzer [5], semi-supervised transfer learning is the case of having abundant labeled source data and limited labeled target data, and unsupervised transfer learning is the case of abundant labeled source data and no labeled target data.…”

“…The paper by Chattopadhyay [14] proposes two separate solutions both using multiple labeled source domains. The first solution is the conditional probability based multisource domain adaptation (CP-MDA) approach, which is a domain adaptation process based on correcting the conditional distribution differences between the source and target domains.…”

On Handling Negative Transfer and Imbalanced Distributions in Multiple Source Transfer Learning

“…Two recently proposed MSTL methods MDA [2] and LWE [9] both weigh each source domain based on the smoothness assumption that a source will gain a high weight if its predictions are smooth among data samples that are close in the feature space. MDA [2] computes a single weight for each source while LWE [9] computes a weight of each source on each sample.…”

“…Thus, we should assign a low weight to domain s in Region R3. This assumption considers local weights for each source, which is different from the global weight assumption held by CRC [10], MDA [2] and GCM [8]. Although LWE [9] calculates local weights, it only considers unsupervised manifolds, which may lead to wrong predictions in negative transfer and imbalanced situations.…”

Multi-Transfer: Transfer Learning with Multiple Views and Multiple Sources

“…For example, the work in [17] extends TrAdaboost [5] by adding a wrapper boosting framework on weighting each source domain. [3] presents a linear combination over multiple sources to reach a consensus. However, these approaches work under singleview setting.…”