Analyzing features for activity recognition

By Tâm Huynh, Bernt Schiele. sOc-EUSAI '05, 2005

377 citations recorded by the source corpus.

Subfields: Computer Vision and Pattern Recognition, Electrical and Electronic Engineering, Transportation

View Paper: https://www.semanticscholar.org/paper/308b57a90f7dc66e8e3efe8bfb9e3ed90b9d4676

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Computer Vision and Pattern Recognition

1. Andreas Bulling
2. Agata Brajdic
3. Didier Stricker
4. Attila Reiss
5. John D. Lafferty
6. Xiaoli Li
7. Manuela M. Veloso
8. Lina Yao

In Electrical and Electronic Engineering

1. Agata Brajdic
2. Stephen J. Preece
3. David Howard
4. John Yannis Goulermas
5. Laurence P. J. Kenney
6. Robert K. Harle
7. John D. Lafferty
8. Manuela M. Veloso

In Transportation

1. Agata Brajdic
2. Robert K. Harle
3. Stephen J. Preece
4. David Howard
5. John Yannis Goulermas
6. Laurence P. J. Kenney
7. John D. Lafferty
8. Manuela M. Veloso

Citations in the corpus

• Classification Accuracies of Physical Activities Using Smartphone Motion Sensors (2012, methodology): “As discussed above, different window sizes have been used in the literature, including 1 second [3,9,15], 2 seconds [20], and 10 seconds [11].”
• Complex Human Activity Recognition Using Smartphone and Wrist-Worn Motion Sensors (2016, background): “Though the effect of window size on simple physical activities has been studied before [13], its effect on complex activities is yet to be explored.”
• Window Size Impact in Human Activity Recognition (2014, background): “Huynh and Schiele (2005) (2 subjects) [43] Walking, jogging, hopping, skipping and more Shoulder strap (1) NCC (∼80%) 0.”
• Preprocessing techniques for context recognition from accelerometer data (2010, result): “In this work, however, we did not explore the impact of window size and of the degree of overlap, as has been studied by other authors [17].”
• Walk detection and step counting on unconstrained smartphones (2013, methodology): “We did not use individual FFT coefficients as under unconstrained movements dominant frequencies tend to vary, similarly as across different types of ambulation [20].”
• Deep Learning for Sensor-based Human Activity Recognition (2020, methodology): “ity Recognition. Numerous previous works adopted machine learning methods in human activity recognition [79]. They highly rely on feature extraction techniques including time-frequency transformation [62], statistical approaches [21] and symbolic representation [88]. However, the features extracted ”
• A Comparison of Feature Extraction Methods for the Classification of Dynamic Activities From Accelerometer Data (2009, methodology): “Both hopping [37] and jumping [36] have been used in previous activity monitoring studies and are also used in”
• A tutorial on human activity recognition using body-worn inertial sensors (2014, background): “…2009] Accelerometers or gyroscopes EOB Detection of body movement patterns, object use, ambient infrastructure [Godfrey et al. 2008; Westeynet al. 2005; Huynh and Schiele 2005; Blanke and Schiele 2009; B ¨ achlin et al. 2009] Magnetometer B Orientation of the body [Lee and Mase 2002] or relative…”
• Deep Convolutional Neural Networks on Multichannel Time Series for Human Activity Recognition (2015, methodology): “By contrast, the traditional features extraction methods such as basis transform coding (e.g. signals with wavelet transform and Fourier transform) [Huynh and Schiele, 2005] , statistics of raw signals (e.g, mean and covariance of time sequences) [Bulling et al., 2014] and symbolic representation…”
• Conditional random fields for activity recognition (2007, methodology): “Huynh and Schiele used features such as fast Fourier transform coefficients computed over windows of accelerometer data to recognize human activities such as walking or riding the bus [41].”
• Creating and benchmarking a new dataset for physical activity monitoring (2012, background): “in [7]), that for segmentation there is no single best window length for all activities.”
• Activity identification using body-mounted sensors—a review of classification techniques (2009, methodology): “This approach was used by Pirttikangas et al (2006) to identify the best sensors/features for the classification 17 different activities Huynh and Schiele (2005) compared a range of different acceleration-derived features, including mean, variance, spectral energy and FFT coefficients.”
• Activity Recognition Using Inertial Sensing for Healthcare, Wellbeing and Sports Applications: A Survey (2010, methodology): “Spectral Energy Energy feature can be used to capture data periodicity of the acceleration data in the frequency domain and it can be used to distinguish sedentary activities from vigorous activities[42].”