Resource-Efficient Pet Dog Sound Events Classification Using LSTM-FCN Based on Time-Series Data

Kim, Yunbin; Sa, Jaewon; Chung, Yongwha; Park, Jooyoung; Lee, Sung-Ju

doi:10.3390/s18114019

Cited by 21 publications

(7 citation statements)

References 30 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Finally, LSTM-FCN has already been deployed in real world scenarios. One such application is to efficiently classify pet dog sounds using resource constrained sensors [28]. The original models, LSTM-FCN and ALSTM-FCN, lacked the explanation of each sub-module.…”

Section: Introductionmentioning

confidence: 99%

Insights Into LSTM Fully Convolutional Networks for Time Series Classification

2019

View full text Add to dashboard Cite

Long Short Term Memory Fully Convolutional Neural Networks (LSTM-FCN) and Attention LSTM-FCN (ALSTM-FCN) have shown to achieve state-of-the-art performance on the task of classifying time series signals on the old University of California-Riverside (UCR) time series repository. However, there has been no study on why LSTM-FCN and ALSTM-FCN perform well. In this paper, we perform a series of ablation tests (3627 experiments) on LSTM-FCN and ALSTM-FCN to provide a better understanding of the model and each of its sub-module. Results from the ablation tests on ALSTM-FCN and LSTM-FCN show that the LSTM and the FCN blocks perform better when applied in a conjoined manner. Two z-normalizing techniques, z-normalizing each sample independently and z-normalizing the whole dataset, are compared using a Wilcoxson signed-rank test to show a statistical difference in performance. In addition, we provide an understanding of the impact dimension shuffle has on LSTM-FCN by comparing its performance with LSTM-FCN when no dimension shuffle is applied. Finally, we demonstrate the performance of the LSTM-FCN when the LSTM block is replaced by a GRU, basic RNN, and Dense Block.

show abstract

Section: Introductionmentioning

confidence: 99%

Insights Into LSTM Fully Convolutional Networks for Time Series Classification

2019

View full text Add to dashboard Cite

show abstract

“…This allows for the analysis of the information and the traceability of the proposal. Data visualization and access are also accessible through the GRAFANA viewer, offering visualization on different time scales and access to data as a flat file for external analysis from any application [58].…”

Section: Information Managementmentioning

confidence: 99%

Optimizing Ambiance: Intelligent RGB Lighting Control in Structures Using Fuzzy Logic

Alvarez-Garcia,

Zúñiga-Cañón,

Garcia-Sanchez

et al. 2024

Applied Sciences

View full text Add to dashboard Cite

Managing red–green–blue (RGB) lighting conditions within structures may evoke emotions and positively influence behavior. Intelligent RGB lighting systems based on environmental data measurements can substantially enhance the perception of comfort. This study presents a challenge that requires a holistic and integrated approach to implement an automatic RGB artificial lighting control system that can be utilized in various structures and indoor environments. Initially, the challenge spans the identification of environmental variables directly impacting comfort up to the careful selection of suitable sensors. The result is the development of a sophisticated and autonomous system that can adjust RGB lighting in real time, creating environments that are both comfortable and energy-efficient. This automated system fosters the creation of appropriate atmospheres across different contexts. The identification and monitoring of environmental variables are achieved through a neuro-fuzzy control mechanism, where fuzzy rules and membership functions are defined based on late positive potential timings and the influence of artificial lighting on human emotions. The outcomes from this study are an interconnected system capable of performing both online and offline operations to enable the monitoring of environmental variables and the efficient management of artificial lighting based on these metrics. A pilot study, with reference to an EEG wave registry system, yielded significant results. These tests had a statistically relevant result with an average frequency of approximately 9.8 Hz, indicative of a state of comfort among people. Despite a 10% deviation margin, 87% of measurements during the test remained consistent. This research study contributes to human behavior by fostering a relaxing environment and enabling a reduction in energy consumption through the use of efficient lighting. Moreover, the environment intention enables the creation of stimuli in three emotional states: activation, relaxation, and neutral, allowing behavioral adaptation to an intention to occur automatically in fluctuating environmental conditions.

show abstract

“…A bidirectional recurrent neural network (BRNN) was first proposed by M Schuster [44]. In several areas, such as phoneme classification [45], speech recognition [46], and emotion classification [47], bidirectional networks outperform unidirectional networks [48]. While applying to the time-series data, it also passes information backward in time and passes in normal temporal sequences.…”

Section: Bidirectional Long Short-term Memory (Blstm)mentioning

confidence: 99%

A Study on Sensor System Latency in VR Motion Sickness

Kundu

Rahman

Paul

2021

JSAN

View full text Add to dashboard Cite

One of the most frequent technical factors affecting Virtual Reality (VR) performance and causing motion sickness is system latency. In this paper, we adopted predictive algorithms (i.e., Dead Reckoning, Kalman Filtering, and Deep Learning algorithms) to reduce the system latency. Cubic, quadratic, and linear functions are used to predict and curve fitting for the Dead Reckoning and Kalman Filtering algorithms. We propose a time series-based LSTM (long short-term memory), Bidirectional LSTM, and Convolutional LSTM to predict the head and body motion and reduce the motion to photon latency in VR devices. The error between the predicted data and the actual data is compared for statistical methods and deep learning techniques. The Kalman Filtering method is suitable for predicting since it is quicker to predict; however, the error is relatively high. However, the error property is good for the Dead Reckoning algorithm, even though the curve fitting is not satisfactory compared to Kalman Filtering. To overcome this poor performance, we adopted deep-learning-based LSTM for prediction. The LSTM showed improved performance when compared to the Dead Reckoning and Kalman Filtering algorithm. The simulation results suggest that the deep learning techniques outperformed the statistical methods in terms of error comparison. Overall, Convolutional LSTM outperformed the other deep learning techniques (much better than LSTM and Bidirectional LSTM) in terms of error.

show abstract

Resource-Efficient Pet Dog Sound Events Classification Using LSTM-FCN Based on Time-Series Data

Cited by 21 publications

References 30 publications

Insights Into LSTM Fully Convolutional Networks for Time Series Classification

Insights Into LSTM Fully Convolutional Networks for Time Series Classification

Optimizing Ambiance: Intelligent RGB Lighting Control in Structures Using Fuzzy Logic

A Study on Sensor System Latency in VR Motion Sickness

Contact Info

Product

Resources

About