Deep Learning Approaches for Fall Detection Using Acoustic Information
01 Jan 2021-pp 479-488
TL;DR: An alert system to be designed to alert the caretaker about the occurrence of the accident in senior citizens' homes by mounting three aerial microphones and one-floor acoustic sensor in their room and monitoring the acoustic information received from the microphone and FAS.
Abstract: Senior citizens are prone to accidents due to their old age. The accidents may cause severe injuries and even to death if it is not identified and treated within a short period of time. Also, it is more risk if they stay alone in their homes. To mitigate the risk, an alert system is to be designed to alert the caretaker about the occurrence of the accident. By mounting three aerial microphones and one-floor acoustic sensor (FAS) in their room and monitoring the acoustic information received from the microphone and FAS, the acoustic information of the fall event is recorded. The acoustic features such as energy, spectral centroid, spectral flux, zero-crossing rate and Mel-frequency cepstral coefficients (MFCC) are extracted from the acoustic signal. Support vector machine (SVM) network and deep learning neural networks (DNN) with more than two hidden layers are trained with a reduced set of features obtained with principal component analysis (PCA) from the acoustic features. DNN classifier is proved to be better than SVM classifier. The obtained accuracy for DNN is 97%, the accuracy of the SVM classifier with MLP kernel and RBF kernel is 50% and 83%, respectively.
Citations
More filters
TL;DR: Wang et al. as discussed by the authors proposed an algorithm that can effectively distinguish falls from a large amount of activities of daily life (ADL) signals, and the proposed method can achieve the highest score in multiple evaluation methods, with a sensitivity of 99.33%, a specificity of 91.86, an F-score of 98.44% and an AUC of98.35%.
Abstract: Falling represents one of the most serious health risks for elderly people; it may cause irreversible injuries if the individual cannot obtain timely treatment after the fall happens. Therefore, timely and accurate fall detection algorithm research is extremely important. Recently, a number of researchers have focused on fall detection and made many achievements, and most of the relevant algorithm studies are based on ideal class-balanced datasets. However, in real-life applications, the possibilities of Activities of Daily Life (ADL) and fall events are different, so the data collected by wearable sensors suffers from class imbalance. The previously developed algorithms perform poorly on class-imbalanced data. In order to solve this problem, this paper proposes an algorithm that can effectively distinguish falls from a large amount of ADL signals. Compared with the state-of-the-art fall detection algorithms, the proposed method can achieve the highest score in multiple evaluation methods, with a sensitivity of 99.33%, a specificity of 91.86%, an F-Score of 98.44% and an AUC of 98.35%. The results prove that the proposed algorithm is effective on class-imbalanced data and more suitable for real-life application compared to previous works.
7 citations
TL;DR: In this paper , a novel feature space mean absolute deviated-local ternary patterns (MAD-LTP) was proposed to examine the environmental sounds and used these features to train the BiLSTM for fall events detection.
3 citations
TL;DR: In this article , an optimized BP neural network fall prediction model based on the Sparrow Search Algorithm (SSA) is established to reduce the injury caused by the fall and solve the problems of low efficiency and low accuracy of traditional fall prediction methods.
2 citations
TL;DR: This work develops a solution based on a Transformer architecture that takes noisy sound input from bathrooms and classifies it into fall/no-fall class with an accuracy of 0.8673, suitable for deploying in elderly homes, hospitals, and rehabilitation facilities without requiring the user to wear any device or be constantly "watched" by the sensors.
2 citations
References
More filters
TL;DR: The proposed approach has a considerable advantage over aerial solutions since it is able to achieve higher fall classification performance using a simpler algorithmic pipeline and hardware setup.
Abstract: An innovative floor acoustic sensor (FAS) for fall classification is presented.A classifier based on Support Vector Machine based fall classifier is developed.A dataset of fall events acquired with FAS and with aerial microphones is described.The FAS rejects high frequency disturbances and that propagate through the air.Classification performance in clean and noisy conditions show the FAS superiority. The interest in assistive technologies for supporting people at home is constantly increasing, both in academia and industry. In this context, the authors propose a fall classification system based on an innovative acoustic sensor that operates similarly to stethoscopes and captures the acoustic waves transmitted through the floor. The sensor is designed to minimize the impact of aerial sounds in recordings, thus allowing a more focused acoustic description of fall events. The audio signals acquired by means of the sensor are processed by a fall recognition algorithm based on Mel-Frequency Cepstral Coefficients, Supervectors and Support Vector Machines to discriminate among different types of fall events. The performance of the algorithm has been evaluated against a specific audio corpus comprising falls of a human mimicking doll and of everyday objects. The results showed that the floor sensor significantly improves the performance respect to an aerial microphone: in particular, the F1-Measure is 6.50% higher in clean conditions and 8.76% higher in mismatched noisy conditions. The proposed approach, thus, has a considerable advantage over aerial solutions since it is able to achieve higher fall classification performance using a simpler algorithmic pipeline and hardware setup.
38 citations
27 Jun 2017
TL;DR: Here, machine learning techniques are applied to process sound simulated the most common type of fall for the elderly, i.e., when the foot collides with an obstacle and the trunk hits the ground before using his/her hands to absorb the fall.
Abstract: One of the most notable consequences of aging is the loss of motor function abilities, making elderly people specially susceptible to falls, which is of the most remarkable concerns in elder care Thus, several solutions have been proposed to detect falls, however, none of them achieved a great success mainly because of the need of wearing a recording device In this paper, we study the use of sound to detect fall events The advantage of this approach over the traditional ones is that the subject does not require to wear additional devices to monitor his or her activities Here, we apply machine learning techniques to process sound simulated the most common type of fall for the elderly, ie, when the foot collides with an obstacle and the trunk hits the ground before using his/her hands to absorb the fall The results show that high levels of accuracy can be achieved using only a few signal processing techniques
10 citations