Analysis of human behavior recognition algorithms based on acceleration data

Bruno, Barbara; Mastrogiovanni, Fulvio; Sgorbissa, Antonio; Vernazza, Tullio; Zaccaria, Renato

doi:10.1109/icra.2013.6630784

Cited by 103 publications

(65 citation statements)

References 14 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Table 2 details the three datasets we use and specifies the classes of behavior they contain. Two of the datasets are audio-based (for speaker identification and emotion recognition) provided by the authors of [25] and one is accelerometer-based [10] containing a general set of Activities of Daily Life (ADL) shown in Table 1. The ADL dataset is composed of the labeled recordings of 14 simple activities performed by 16 volunteers wearing a single tri-axial accelerometer attached to the right wrist of the volunteer and sampled at a rate of 32Hz.…”

Section: Experiments Setupmentioning

confidence: 99%

Can Deep Learning Revolutionize Mobile Sensing?

Lane

Georgiev

2015

Proceedings of the 16th International Workshop on Mobile Computing Systems and Applications

219

105

View full text Add to dashboard Cite

Sensor-equipped smartphones and wearables are transforming a variety of mobile apps ranging from health monitoring to digital assistants. However, reliably inferring user behavior and context from noisy and complex sensor data collected under mobile device constraints remains an open problem, and a key bottleneck to sensor app development. In recent years, advances in the field of deep learning have resulted in nearly unprecedented gains in related inference tasks such as speech and object recognition. However, although mobile sensing shares many of the same data modeling challenges, we have yet to see deep learning be systematically studied within the sensing domain. If deep learning could lead to significantly more robust and efficient mobile sensor inference it would revolutionize the field by rapidly expanding the number of sensor apps ready for mainstream usage.In this paper, we provide preliminary answers to this potentially game-changing question by prototyping a low-power Deep Neural Network (DNN) inference engine that exploits both the CPU and DSP of a mobile device SoC. We use this engine to study typical mobile sensing tasks (e.g., activity recognition) using DNNs, and compare results to learning techniques in more common usage. Our early findings provide illustrative examples of DNN usage that do not overburden modern mobile hardware, while also indicating how they can improve inference accuracy. Moreover, we show DNNs can gracefully scale to larger numbers of inference classes and can be flexibly partitioned across mobile and remote resources. Collectively, these results highlight the critical need for further exploration as to how the field of mobile sensing can best make use of advances in deep learning towards robust and efficient sensor inference.

show abstract

Section: Experiments Setupmentioning

confidence: 99%

Can Deep Learning Revolutionize Mobile Sensing?

Lane

Georgiev

2015

Proceedings of the 16th International Workshop on Mobile Computing Systems and Applications

219

105

View full text Add to dashboard Cite

show abstract

“…In order to compare randomization types, we use a wristworn 3-axis accelerometer dataset consisting of the execution of 7 motion primitives by 10 different volunteers for 10 times [7]. Fig.…”

Section: A Dataset and Feature Extractionmentioning

confidence: 99%

Source-Aware Partitioning for Robust Cross-Validation

Kilinc

Uysal

2015

2015 IEEE 14th International Conference on Machine Learning and Applications (ICMLA)

View full text Add to dashboard Cite

One of the most critical components of engineering a machine learning algorithm for a live application is robust performance assessment prior to its implementation.Crossvalidation is used to forecast a specific algorithm's classification or prediction accuracy on new input data given a finite dataset for training and testing the algorithm. Two most well known cross-validation techniques, random subsampling (RSS) and Kfold, are used to generalize the assessment results of machine learning algorithms in a non-exhaustive random manner. In this work we first show that for an inertia based activity recognition problem where data is collected from different users of a wristworn wireless accelerometer, random partitioning of the data, regardless of cross-validation technique, results in statistically similar average accuracies for a standard feed-forward neural network classifier. We propose a novel source-aware partitioning technique where samples from specific users are completely left out of the training/validation sets in rotation. The average error for the proposed cross-validation method is significantly higher with lower standard variation, which is a major indicator of cross-validation robustness. Approximately 30% increase in average error rate implies that source-aware cross validation could be a better indication of live algorithm performance where test data statistics would be significantly different than training data due to source (or user)-sensitive nature of process data.

show abstract

“…Many new software of today require real-time monitoring of human activity to perform their task [16,17,19,20,21,22]. Sensors are attached to a human subject.…”

Section: Introductionmentioning

confidence: 99%

“…They send streams of real-time time-series data for analysis in order for computer programs to formulate a proper response. Of great importance is to figure out what kind of activity a human subject is engaged in based on time-series data the computer is receiving [21,22,23]. Each activity shows distinctive is ideally suited to performing such a transform.…”

Section: Introductionmentioning

confidence: 99%

Applying Hilbert-Huang Transform to Extract Essential Patterns from Hand Accelerometer Data

Choe¹,

Suh²

2017

JIIBC

View full text Add to dashboard Cite

Hand Accelerometers are widely used to detect human motion patterns in real-time. It is essential to reliably identify which type of activity is performed by human subjects. This rests on having accurate template of each activity. Many human activities are represented as a set of multiple time-series data from such sensors, which are mostly non-stationary and non-linear in nature. This requires a method which can effectively extract patterns from non-stationary and non-linear data. To achieve such a goal, we propose the method to apply Hilbert-Huang Transform which is known to be an effective way of extracting non-stationary and non-linear components from time-series data. It is applied on samples of accelerometer data to determine its effectiveness.

show abstract

Analysis of human behavior recognition algorithms based on acceleration data

Cited by 103 publications

References 14 publications

Can Deep Learning Revolutionize Mobile Sensing?

Can Deep Learning Revolutionize Mobile Sensing?

Source-Aware Partitioning for Robust Cross-Validation

Applying Hilbert-Huang Transform to Extract Essential Patterns from Hand Accelerometer Data

Contact Info

Product

Resources

About