Indirect Recognition of Predefined Human Activities

Abstract

The work investigates the application of artificial neural networks and logistic regression for the recognition of activities performed by room occupants. KNX (Konnex) standard-based devices were selected for smart home automation and data collection. The obtained data from these devices (Humidity, CO2, temperature) were used in combination with two wearable gadgets to classify specific activities performed by the room occupant. The obtained classifications can benefit the occupant by monitoring the wellbeing of elderly residents and providing optimal air quality and temperature by utilizing heating, ventilation, and air conditioning control. The obtained results yield accurate classification.

Human Activity Classification Using Multilayer Perceptron

“…This allows better training of larger models in a given time. With direct comparison with our previous study [ 13 ], which used two hidden layers, the cross-validation accuracy was almost identical (within margins of error). However, due to higher and more stable data acquisition rates, the scoring accuracy was significantly improved.…”

“…A total of 84 models were developed to examine the recognition accuracy of these activity classes. The models used for cross-validation (42 models) mostly showed accuracy levels above 99%, which is considerably more accurate than similar implementations and our previous study [ 13 ]. The relaxing activity showed mostly 100% recognition accuracy levels, and other activities cross-validated to accuracy levels above 98.86%.…”

Human Activity Classification Using Multilayer Perceptron

“…This allows better training of larger models in a given time. With direct comparison with our previous study [ 13 ], which used two hidden layers, the cross-validation accuracy was almost identical (within margins of error). However, due to higher and more stable data acquisition rates, the scoring accuracy was significantly improved.…”

“…A total of 84 models were developed to examine the recognition accuracy of these activity classes. The models used for cross-validation (42 models) mostly showed accuracy levels above 99%, which is considerably more accurate than similar implementations and our previous study [ 13 ]. The relaxing activity showed mostly 100% recognition accuracy levels, and other activities cross-validated to accuracy levels above 98.86%.…”

Using Smart Home Technologies to Promote Physical Activity Among the General and Aging Populations: Scoping Review

“…Other algorithms such as artificial neural networks (ANNs), convolutional neural networks, and hidden Markov models were only applied once. Compared with Bayesian or neural networks, which require and use a large amount of training data sets (eg, the study by Gorjani et al [ 35 ], which used approximately 300,000 data points), support vector machine used a small number of available training sets (eg, the study by Li et al [ 39 ], which used <200 data points).…”

“…The included studies were conducted in various settings. In total, 40% (8/20) were conducted in a laboratory setting [ 32 , 35 , 37 , 39 - 41 , 48 , 49 ], 25% (5/20) were conducted in a home setting [ 3 , 32 , 33 , 43 , 51 ], and 10% (2/20) were conducted in a hospital setting [ 42 , 47 ]. Laboratory setting means that the study was conducted in an environment where researchers installed smart home sensors in a laboratory for data collection instead of in a real home setting, which has basic rooms and equipment for daily living.…”

A Cervical Plethysmography System to Monitor Blood Vessel Pulses on the Neck

“…Moreover, with the proposed system it is possible to differentiate dissimilar activities. Indeed, when the chair was tilted backward from 90 • to 0 • , the acquired signals had large peaks (Figure 1A), which can be analyzed with machine learning techniques for activity recognition purposes [8].…”

Human Activity Classification Using Multilayer Perceptron

“…This allows better training of larger models in a given time. With direct comparison with our previous study [ 13 ], which used two hidden layers, the cross-validation accuracy was almost identical (within margins of error). However, due to higher and more stable data acquisition rates, the scoring accuracy was significantly improved.…”

“…A total of 84 models were developed to examine the recognition accuracy of these activity classes. The models used for cross-validation (42 models) mostly showed accuracy levels above 99%, which is considerably more accurate than similar implementations and our previous study [ 13 ]. The relaxing activity showed mostly 100% recognition accuracy levels, and other activities cross-validated to accuracy levels above 98.86%.…”

Using Smart Home Technologies to Promote Physical Activity Among the General and Aging Populations: Scoping Review

“…Other algorithms such as artificial neural networks (ANNs), convolutional neural networks, and hidden Markov models were only applied once. Compared with Bayesian or neural networks, which require and use a large amount of training data sets (eg, the study by Gorjani et al [ 35 ], which used approximately 300,000 data points), support vector machine used a small number of available training sets (eg, the study by Li et al [ 39 ], which used <200 data points).…”

“…The included studies were conducted in various settings. In total, 40% (8/20) were conducted in a laboratory setting [ 32 , 35 , 37 , 39 - 41 , 48 , 49 ], 25% (5/20) were conducted in a home setting [ 3 , 32 , 33 , 43 , 51 ], and 10% (2/20) were conducted in a hospital setting [ 42 , 47 ]. Laboratory setting means that the study was conducted in an environment where researchers installed smart home sensors in a laboratory for data collection instead of in a real home setting, which has basic rooms and equipment for daily living.…”

A Cervical Plethysmography System to Monitor Blood Vessel Pulses on the Neck

“…Moreover, with the proposed system it is possible to differentiate dissimilar activities. Indeed, when the chair was tilted backward from 90 • to 0 • , the acquired signals had large peaks (Figure 1A), which can be analyzed with machine learning techniques for activity recognition purposes [8].…”

Human Activity Classification Using Multilayer Perceptron

“…This allows better training of larger models in a given time. With direct comparison with our previous study [ 13 ], which used two hidden layers, the cross-validation accuracy was almost identical (within margins of error). However, due to higher and more stable data acquisition rates, the scoring accuracy was significantly improved.…”

“…A total of 84 models were developed to examine the recognition accuracy of these activity classes. The models used for cross-validation (42 models) mostly showed accuracy levels above 99%, which is considerably more accurate than similar implementations and our previous study [ 13 ]. The relaxing activity showed mostly 100% recognition accuracy levels, and other activities cross-validated to accuracy levels above 98.86%.…”

Using Smart Home Technologies to Promote Physical Activity Among the General and Aging Populations: Scoping Review

“…Other algorithms such as artificial neural networks (ANNs), convolutional neural networks, and hidden Markov models were only applied once. Compared with Bayesian or neural networks, which require and use a large amount of training data sets (eg, the study by Gorjani et al [ 35 ], which used approximately 300,000 data points), support vector machine used a small number of available training sets (eg, the study by Li et al [ 39 ], which used <200 data points).…”

“…The included studies were conducted in various settings. In total, 40% (8/20) were conducted in a laboratory setting [ 32 , 35 , 37 , 39 - 41 , 48 , 49 ], 25% (5/20) were conducted in a home setting [ 3 , 32 , 33 , 43 , 51 ], and 10% (2/20) were conducted in a hospital setting [ 42 , 47 ]. Laboratory setting means that the study was conducted in an environment where researchers installed smart home sensors in a laboratory for data collection instead of in a real home setting, which has basic rooms and equipment for daily living.…”

A Cervical Plethysmography System to Monitor Blood Vessel Pulses on the Neck

“…Moreover, with the proposed system it is possible to differentiate dissimilar activities. Indeed, when the chair was tilted backward from 90 • to 0 • , the acquired signals had large peaks (Figure 1A), which can be analyzed with machine learning techniques for activity recognition purposes [8].…”