On-bed sleep posture recognition based on body-earth mover's distance

Xu, Xiaowei; Lin, Feng; Wang, Aosen; Song, Chen; Hu, Yu; Xu, Wenyao

doi:10.1109/biocas.2015.7348281

Cited by 22 publications

(7 citation statements)

References 12 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Thus, HEMD adopts principal component analysis (PCA) to estimate the angle of the major axis of variation and to rotate the image according to the variation. The rotation angel is equal to the angle between the first eigenvectors of the PCA results and the edge of the image [15].…”

Section: Heat-earth Mover's Distancementioning

confidence: 99%

Thermal handprint analysis for forensic identification using Heat-Earth Mover's Distance

Cho

Lin

Song

et al. 2016

2016 IEEE International Conference on Identity, Security and Behavior Analysis (ISBA)

Self Cite

View full text Add to dashboard Cite

In recent years, handprint-based recognition system has been widely applied in forensic science. However, existing identification systems are nearly based on the handprints that could be easily prevented. In contrast to earlier works, we exploit the thermal handprint dissimilarity measure, called Heat-Earth Mover's Distance (HEMD). It is designed to classify heat-based handprints that can be obtained even when the subject wears a glove.Our objective is to develop a system that performs a forensic identification using heat-based handprints 1. Thermal handprints made without the glove 2. Thermal handprint made with the glove Existing forensic identification systems:1. DNA fingerprinting (blood, skin, hair, saliva) uses a length of the strands of the DNA molecules with repeating base pair patterns by restricting fragment length polymorphism (RFLP) 2. Odor analysis uses a chemical analysis of perfume components by gas chromatography (GC) 3. Voice analysis using Gaussian mixture speaker models achieved an accuracy of 82.8% (Reynolds et al). 4. Fingerprint / handprint uses a minutiae-based structural matching. Previously, handprint-based verification system made by integrating hand geometry feature and fingerprint feature achieved an accuracy of 97% (Yan et al). when acceptance error and rejection error become equal. The lower the EER, the higher the accuracy. Relatively low EER of SH indicates that SH is more stable than using GH Figure 5. our task is to transport the points from the target X to training Y with the least cost. Second equation allows the overall weight of points in X to be equal to the overall weight of points in Y. Thus, weight normalization is not needed. Figure 2.Limitations:

show abstract

Section: Heat-earth Mover's Distancementioning

confidence: 99%

Thermal handprint analysis for forensic identification using Heat-Earth Mover's Distance

Cho

Lin

Song

et al. 2016

2016 IEEE International Conference on Identity, Security and Behavior Analysis (ISBA)

Self Cite

View full text Add to dashboard Cite

show abstract

“…Using the resulting interquartile ranges for the arc lengths at each frequency the efficacy of the decision algorithms was significantly increased. Using TA2 = 1.25 for 2.4 GHz, there was three erroneous decision (subject 6, 7 and 18) in prone and side posture assessment and using TA2 = 0.53 for 5.8 GHz, there were in total six erroneous decisions (subjects 6,7,9,11,17,20) for supine, prone and side posture assessment of twenty subjects. Here, it should be mentioned that new thresholds were fit on the twenty subjects dataset whereas the old threshold was applicable for 18 subjects.…”

Section: Posture Recognitionmentioning

confidence: 99%

“…One of the most common sleep disorders is sleep apnea, characterized by pauses in breathing during sleep that can increase the potential risk of serious diseases such as hypertension and bipolar disorder [1,3,4]. Various clinical investigations have demonstrated that the recoginition of key categories of posture assumed during sleep can serve as a diagnostic indicator for a variety of chronic diseases and can potentially aid in medical therapies [5][6][7][8][9]. The predominant categories of concern in sleep medicine are supine, prone, and side.…”

Section: Introductionmentioning

confidence: 99%

“…However, this approach is contact based, requiring specialized instruments and trained professionals that can make the procedure too cumbersome and expensive for many applications. Video cameras have also been used to recognize sleep postures [14], however, the collection of video imagery introduces privacy issues, and the accuracy is hampered by obstructions such as bed covers [6][7][8][9]. Pressure-sensitive bedsheets with densely deployed textile sensors have been used to generate pressure map images, with machine learning (ML) classifiers used to recognize sleep postures [15].…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Sleep Posture Recognition With a Dual-Frequency Cardiopulmonary Doppler Radar

et al. 2021

View full text Add to dashboard Cite

While Doppler radar can be used to measure cardiopulmonary vital signs during sleep, meaningful diagnostic assessments are often subject to knowledge of a subject's changing sleep posture. The torso Effective Radar Cross Section (ERCS) and displacement magnitude were studied for 20 human subjects in three imitated sleep posture categories using a dual-frequency Doppler radar system in an exploratory examination of the feasibility of using radar to recognize body orientation. Box plot statistical analyses were performed for comparative assessment of ratio variations in ERCS and respiration depth for three different imitated sleep postures. The observed statistical trends and correlations were applied to a physical model to develop posture decision algorithms with initial supine posture data used as a reference. A single-frequency algorithm tracked postures without error for 90% of the subjects using 2.4 GHz data, and 80% using 5.8 GHz data. As accuracy limitations were complementary, a dual-frequency algorithm was developed which recognized postures without error for 100% of the subjects.

show abstract

“…Xu et al proposed a new distance measurement method for sleep posture differencing. By projecting the pressure distribution to the horizontal and vertical directions, distribution differences can be identified and it achieved 90.78% high accuracy through a KNN classifier [11]. They improved the model and proposed a novel matching-based approach in the next year achieving 91.21% accuracy [12].…”

Section: A Sleep Posture Recognitionmentioning

confidence: 99%

Self-Supervised Learning From Multi-Sensor Data for Sleep Recognition

2020

View full text Add to dashboard Cite

Sleep recognition refers to detection or identification of sleep posture, state or stage, which can provide critical information for the diagnosis of sleep diseases. Most of sleep recognition methods are limited to single-task recognition, which only involves single-modal sleep data, and there is no generalized model for multi-task recognition on multi-sensor sleep data. Moreover, the shortage and imbalance of sleep samples also limits the expansion of the existing machine learning methods like support vector machine, decision tree and convolutional neural network, which lead to the decline of the learning ability and overfitting. Self-supervised learning technologies have shown their capabilities to learn significant feature representations. In this paper, a novel self-supervised learning model is proposed for sleep recognition, which is composed of an upstream self-supervised pre-training task and a downstream recognition task. The upstream task is conducted to increase the data capacity, and the information of frequency domain and the rotation view are used to learn the multi-dimensional sleep feature representations. The downstream task is undertaken to fuse bidirectional long-short term memory and conditional random field as the sequential data recognizer to produce the sleep labels. Our experiments shows that our proposed algorithm provide promising results in sleep identification and can further be applied in clinical and smart home environments as a diagnostic tool. The source code is provided at: ''https://github.com/zhaoaite/SSRM ''.

show abstract

On-bed sleep posture recognition based on body-earth mover's distance

Cited by 22 publications

References 12 publications

Thermal handprint analysis for forensic identification using Heat-Earth Mover's Distance

Thermal handprint analysis for forensic identification using Heat-Earth Mover's Distance

Sleep Posture Recognition With a Dual-Frequency Cardiopulmonary Doppler Radar

Self-Supervised Learning From Multi-Sensor Data for Sleep Recognition

Contact Info

Product

Resources

About