Long Short-term Memory Neural Network Research Articles

Water level prediction of Liuxihe Reservoir based on improved long short-term memory neural network

Abstract To meet the demand of accurate water level prediction of the reservoir in Liuxihe River Basin, this paper proposes an improved long short-term memory (LSTM) neural network based on the Bayesian optimization algorithm and wavelet decomposition coupling. Based on the improved model, the water levels of Liuxihe Reservoir and Huanglongdai Reservoir are simulated and predicted by the 1 h prediction length, and the prediction accuracy of the improved model is verified separately by the 3, 6 and 12 h prediction lengths. The results show that: first, Bayesian optimization coupling can significantly reduce the average absolute error and root mean square error of the model and improve the overall prediction accuracy, but this algorithm is insufficient in the optimization of model extremum; Wavelet decomposition coupling can significantly reduce the outliers in model prediction and improve the accuracy of extremum, but it plays relatively weaker role in the overall optimization of the model. Second, by the prediction lengths of 1, 3, 6 and 12 h, the improved model based on the LSTM neural network and coupled with Bayesian optimization and wavelet decomposition is superior to Bayesian optimization and wavelet decomposition coupling model in overall prediction accuracy and prediction accuracy of extremum.

A Study of Deep Learning Neural Network Algorithms and Genetic Algorithms for FJSP

Flexible job-shop scheduling problem (FJSP) is a new research hotspot in the field of production scheduling. To solve the multiobjective FJSP problem, the production of flexible job shop can run normally and quickly. This research takes into account various characteristics of FJSP problems, such as the need to ensure the continuity and stability of processing, the existence of multiple objectives in the whole process, and the constant complexity of changes. It starts with deep learning neural networks and genetic algorithms. Long short-term memory (LSTM) and convolutional neural networks (CNN) are combined in deep learning neural networks. The new improved algorithm is based on the combination of deep learning neural networks LSTM and CNN with genetic algorithm (GA), namely, CNN-LSTM-GA algorithm. Simulation results showed that the accuracy of the CNN-LSTM-GA algorithm was between 85.2% and 95.3% in the test set. In the verification set, the minimum accuracy of the CNN-LSTM-GA algorithm was 84.6%, both of which were higher than the maximum accuracy of the other two algorithms. In the FJSP simulation experiment, the AUC value of the CNN-LSTM-GA algorithm was 0.92. After 40 iterations, the F1 value of the CNN-LSTM-GA algorithm remained above 0.8, which was significantly higher than the other two algorithms. CNN-LSTM-GA is superior to the other two algorithms in terms of prediction accuracy and overall performance of FJSP. It is more suitable for solving the discrete manufacturing job scheduling problem with FJSP characteristics. This study significantly raises the utilisation rate of the assembly shop’s equipment, optimises the scheduling of FJSP, and fully utilises each processing device’s versatile characteristics, which are quite useful for the production processes of domestic vehicle manufacturing companies.

Advanced Human Activity Recognition through Data Augmentation and Feature Concatenation of Micro-Doppler Signatures

Developing accurate classification models for radar-based Human Activity Recognition (HAR), capable of solving real-world problems, depends heavily on the amount of available data. In this paper, we propose a simple, effective, and generalizable data augmentation strategy along with preprocessing for micro-Doppler signatures to enhance recognition performance. By leveraging the decomposition properties of the Discrete Wavelet Transform (DWT), new samples are generated with distinct characteristics that do not overlap with those of the original samples. The micro-Doppler signatures are projected onto the DWT space for the decomposition process using the Haar wavelet. The returned decomposition components are used in different configurations to generate new data. Three new samples are obtained from a single spectrogram, which increases the amount of training data without creating duplicates. Next, the augmented samples are processed using the Sobel filter. This step allows each sample to be expanded into three representations, including the gradient in the x-direction (Dx), y-direction (Dy), and both x- and y-directions (Dxy). These representations are used as input for training a three-input convolutional neural network-long short-term memory support vector machine (CNN-LSTM-SVM) model. We have assessed the feasibility of our solution by evaluating it on three datasets containing micro-Doppler signatures of human activities, including Frequency Modulated Continuous Wave (FMCW) 77 GHz, FMCW 24 GHz, and Impulse Radio Ultra-Wide Band (IR-UWB) 10 GHz datasets. Several experiments have been carried out to evaluate the model's performance with the inclusion of additional samples. The model was trained from scratch only on the augmented samples and tested on the original samples. Our augmentation approach has been thoroughly evaluated using various metrics, including accuracy, precision, recall, and F1-score. The results demonstrate a substantial improvement in the recognition rate and effectively alleviate the overfitting effect. Accuracies of 96.47%, 94.27%, and 98.18% are obtained for the FMCW 77 GHz, FMCW 24 GHz, and IR- UWB 10 GHz datasets, respectively. The findings of the study demonstrate the utility of DWT to enrich micro-Doppler training samples to improve HAR performance. Furthermore, the processing step was found to be efficient in enhancing the classification accuracy, achieving 96.78%, 96.32%, and 100% for the FMCW 77 GHz, FMCW 24 GHz, and IR-UWB 10 GHz datasets, respectively.

Land Subsidence Prediction Model Based on the Long Short-Term Memory Neural Network Optimized Using the Sparrow Search Algorithm

Land subsidence is a prevalent geological issue that poses significant challenges to construction projects. Consequently, the accurate prediction of land subsidence has emerged as a focal point of research among scholars and experts. Traditional mathematical models exhibited certain limitations in forecasting the extent of land subsidence. To address this issue, the sparrow search algorithm (SSA) was introduced to optimize the efficacy of the long short-term memory (LSTM) neural network in land subsidence prediction. This prediction model has been successfully applied to the Huanglong Commercial City project in the Guanghua unit of Wenzhou city, Zhejiang province, China, and has been compared with the predictions of other models. Using monitoring location 1 as a reference, the MAE, MSE, and RMSE of the test samples for the LSTM neural network optimized using the SSA are 0.0184, 0.0004, and 0.0207, respectively, demonstrating a commendable predictive performance. This new model provides a fresh strategy for the land subsidence prediction of the project and offers new insights for further research on combined models.

PTS-LSTM: Temperature prediction for fused filament fabrication using thermal image time series

In-situ thermography for Fuse Filament Fabrication (FFF) processes reveals the dynamic thermal behavior during printing. The data collected are thermal image time series. Their infrared (IR) intensity is visual evidence of heat-affected zone (HAZ) temperatures, which can be leveraged to train deep learning models, e.g., Long Short-Term Memory (LSTM), for real-time temperature prediction and process monitoring. Nonetheless, the data collection method and printing path may pose challenges for data modeling. Typically, the IR camera has a fixed position while the HAZ moves per the predetermined printing path. Consequently, the HAZ shifts in images and the features extracted from these thermal images show a “periodic” behavior over time. Such periodic patterns do not reflect any useful information about HAZ temperatures. Instead, they are noise hiding and interrupting the true temperature information, thus must be removed before using the data to train an LSTM model for temperature prediction. This study integrates a time series model, i.e., ARIMA, with Stacked LSTM to build a pTS-LSTM model that eliminates noisy patterns and predicts temperatures during FFF printing. The case study results show the outperformance of pTS-LSTM over conventional LSTM and classic Recurrent Neural Network models. pTS-LSTM is demonstrated to be promising for in-situ process monitoring with low-quality thermal images. In FFF practices, pTS-LSTM will be a preferred option over the commonly used deep learning models for thermal-image-based temperature prediction.

Adversarial attack and training for deep neural network based power quality disturbance classification

Power quality disturbance (PQD) can significantly affect the normal operation of the power system. Deep neural network (DNN) can classify PQD with extremely high accuracy. However, adversarial attacks can drastically degrade the performance of DNN, which raises security issues for DNN-based PQD classification. At present, there are few researches on adversarial attacks and their defenses in PQD classification. In this study, we first adopt convolutional neural network-long short-term memory (CNN-LSTM) to classify PQD signals. Then we propose an adversarial attack algorithm for PQD classification, i.e., momentum iterative-fast gradient sign method (MI-FGSM). MI-FGSM generates adversarial perturbations along the gradient direction of the loss function by incorporating momentum during iterations. Finally, we propose a defense algorithm to defend against adversarial attacks, i.e., iterative adversarial training (IAT). IAT can improve the robustness of the classification model by making it learn to minimize the worst loss caused by multi-class adversarial perturbation strengths. The experimental results demonstrate that compared with the fast gradient sign method, the MI-FGSM produces smaller perturbations and can further reduce the classification accuracy of CNN-LSTM. In addition, compared with the most advanced adversarial training defense algorithm in PQD classification, IAT can effectively enhance the robustness of the CNN-LSTM against adversarial attacks.

The importance of interpreting machine learning models for blood glucose prediction in diabetes: an analysis using SHAP

Machine learning has become a popular tool for learning models of complex dynamics from biomedical data. In Type 1 Diabetes (T1D) management, these models are increasingly been integrated in decision support systems (DSS) to forecast glucose levels and provide preventive therapeutic suggestions, like corrective insulin boluses (CIB), accordingly. Typically, models are chosen based on their prediction accuracy. However, since patient safety is a concern in this application, the algorithm should also be physiologically sound and its outcome should be explainable. This paper aims to discuss the importance of using tools to interpret the output of black-box models in T1D management by presenting a case-of-study on the selection of the best prediction algorithm to integrate in a DSS for CIB suggestion. By retrospectively “replaying” real patient data, we show that two long-short term memory neural networks (LSTM) (named p-LSTM and np-LSTM) with similar prediction accuracy could lead to different therapeutic decisions. An analysis with SHAP—a tool for explaining black-box models’ output—unambiguously shows that only p-LSTM learnt the physiological relationship between inputs and glucose prediction, and should therefore be preferred. This is verified by showing that, when embedded in the DSS, only p-LSTM can improve patients’ glycemic control.

Using the LSTM Neural Network and the UWB Positioning System to Predict the Position of Low and High Speed Moving Objects.

Automation of transportation will play a crucial role in the future when people driving vehicles will be replaced by autonomous systems. Currently, the positioning systems are not used alone but are combined in order to create cooperative positioning systems. The ultra-wideband (UWB) system is an excellent alternative to the global positioning system (GPS) in a limited area but has some drawbacks. Despite many advantages of various object positioning systems, none is free from the problem of object displacement during measurement (data acquisition), which affects positioning accuracy. In addition, temporarily missing data from the absolute positioning system can lead to dangerous situations. Moreover, data pre-processing is unavoidable and takes some time, affecting additionally the object's displacement in relation to its previous position and its starting point of the new positioning process. So, the prediction of the position of an object is necessary to minimize the time when the position is unknown or out of date, especially when the object is moving at high speed and the position update rate is low. This article proposes using the long short-term memory (LSTM) artificial neural network to predict objects' positions based on historical data from the UWB system and inertial navigation. The proposed solution creates a reliable positioning system that predicts 10 positions of low and high-speed moving objects with an error below 10 cm. Position prediction allows detection of possible collisions-the intersection of the trajectories of moving objects.

Analyzing influence of COVID-19 on crypto & financial markets and sentiment analysis using deep ensemble model.

COVID-19 affected the world's economy severely and increased the inflation rate in both developed and developing countries. COVID-19 also affected the financial markets and crypto markets significantly, however, some crypto markets flourished and touched their peak during the pandemic era. This study performs an analysis of the impact of COVID-19 on public opinion and sentiments regarding the financial markets and crypto markets. It conducts sentiment analysis on tweets related to financial markets and crypto markets posted during COVID-19 peak days. Using sentiment analysis, it investigates the people's sentiments regarding investment in these markets during COVID-19. In addition, damage analysis in terms of market value is also carried out along with the worse time for financial and crypto markets. For analysis, the data is extracted from Twitter using the SNSscraper library. This study proposes a hybrid model called CNN-LSTM (convolutional neural network-long short-term memory model) for sentiment classification. CNN-LSTM outperforms with 0.89, and 0.92 F1 Scores for crypto and financial markets, respectively. Moreover, topic extraction from the tweets is also performed along with the sentiments related to each topic.

Aquaculture 4.0: hybrid neural network multivariate water quality parameters forecasting model

This study examined the efficiency of hybrid deep neural network and multivariate water quality forecasting model in aquaculture ecosystem. Accurate forecasting of critical water quality parameters can allow for timely identification of possible problem areas and enable decision-makers to take pre-emptive remedial actions that can significantly improve water quality management in aquaculture industry. A novel hybrid deep learning neural network multivariate water quality parameters forecasting model is developed with the aid of ensemble empirical mode decomposition (EEMD) method, deep learning long-short term memory (LSTM) neural network (NN), and multivariate linear regression (MLR) method. The presented water quality forecasting model (shortened as EEMD–MLR–LSTM NN model) is developed using multivariate time-series water quality sensor data collected from Loch Duart company, a Salmon offshore aquaculture farm based around Scourie, northwest Scotland. The performance of the novel hybrid water quality forecasting model is validated by comparing the forecast result with measured water quality parameters data and the real Phytoplankton data count from the aquaculture farm. The forecast accuracy of the results suggests that the novel hybrid water quality forecasting model can be used as a valuable support tool for water quality management in aquaculture industries.

Deep learning-based algorithms for long-term prediction of chlorophyll-a in catchment streams

Accurate estimation of harmful algal blooms is imperative for the protection of surface water. Chlorophyll-a (Chl-a) which is used as a proxy for estimating the algal concentration, is affected by a wide range of weather and physicochemical factors that act at varying spatial and temporal scales. Deep learning (DL) based models such as Long-Short Term Memory (LSTM) and Convolution Neural Networks (CNNs) have shown promising results for hydrological and Chl-a simulations. Recently several variants of LSTM and CNNs have been developed which can model highly non-linear relationships between input and target data. Therefore, these advanced DL methods have the potential for long-term simulation of Chl-a. Previous DL-based studies on Chl-a simulation have developed site-dependent models. This indicates that the DL models were trained and evaluated using data from the same site, making it difficult to apply these models to other sites. Development of site-independent models requires a more robust training strategy which can result in DL models that can be evaluated in new novel situations. To address these issues, we propose a DL-based framework which can incorporate irregularly measured water quality observations, static physical features, and climate data measured at constant time steps. In this framework, we compared the performance of six state of the art DL methods which include (1) LSTM, (2) CNN, (3) Temporal Convolution Networks (TCN), (4) CNN-LSTM, (5) LSTM based autoencoder, and (6) input-attention LSTM (IA-LSTM). The IA-LSTM is an explainable DL method which can select important hydrologic, climatic and water quality parameters for Chl-a prediction. Our results indicate that the IA-LSTM exhibited the best performance, with an R2 value of 0.85 at the training site and 0.52 at the test site. We showed that attention-based deep learning models improve the prediction performance and make the black-box deep learning models interpretable and explainable. The attention-based deep learning models indicated that Chl-a concentration in the Nakdong River was strongly affected by climate factors during the previous three days. The proposed DL framework can be adopted to develop regional water quality models using deep learning.

Secure and Privacy-Preserving Intrusion Detection and Prevention in the Internet of Unmanned Aerial Vehicles.

In smart cities, unmanned aerial vehicles (UAVS) play a vital role in surveillance, monitoring, and data collection. However, the widespread integration of UAVs brings forth a pressing concern: security and privacy vulnerabilities. This study introduces the SP-IoUAV (Secure and Privacy Preserving Intrusion Detection and Prevention for UAVS) model, tailored specifically for the Internet of UAVs ecosystem. The challenge lies in safeguarding UAV operations and ensuring data confidentiality. Our model employs cutting-edge techniques, including federated learning, differential privacy, and secure multi-party computation. These fortify data confidentiality and enhance intrusion detection accuracy. Central to our approach is the integration of deep neural networks (DNNs) like the convolutional neural network-long short-term memory (CNN-LSTM) network, enabling real-time anomaly detection and precise threat identification. This empowers UAVs to make immediate decisions in dynamic environments. To proactively counteract security breaches, we have implemented a real-time decision mechanism triggering alerts and initiating automatic blacklisting. Furthermore, multi-factor authentication (MFA) strengthens access security for the intrusion detection system (IDS) database. The SP-IoUAV model not only establishes a comprehensive machine framework for safeguarding UAV operations but also advocates for secure and privacy-preserving machine learning in UAVS. Our model's effectiveness is validated using the CIC-IDS2017 dataset, and the comparative analysis showcases its superiority over previous approaches like FCL-SBL, RF-RSCV, and RBFNNs, boasting exceptional levels of accuracy (99.98%), precision (99.93%), recall (99.92%), and F-Score (99.92%).

Vibration prediction of offshore wind turbines based on long short-term memory network

ABSTRACT Offshore wind turbines (OWTs) have complex operating conditions and are prone to accidents. Timely and accurate vibration prediction can reduce the probability of failure of OWTs. To address this issue, this study proposed a vibration prediction model based on long short-term memory (LSTM) network. The proposed prediction model was applied to an OWT in an offshore wind farm in China, and three evaluation indicators, i.e. mean absolute error (MAE), mean absolute percentage error (MAPE), and root mean square error (RMSE), were used to verify the vibration prediction performance of LSTM. The results demonstrated that the prediction accuracy of LSTM was better than that of back propagation (BP) neural network, and LSTM had optimum engineering practicability. The prediction accuracy of LSTM and BP neural networks decreased with an increase in prediction durations. However, the prediction accuracy of LSTM was higher. This study improves the online safety monitoring of OWTs.

Seismic response prediction of a train-bridge coupled system based on a LSTM neural network

High complexity and randomness in high-speed train-bridge interactive dynamic analysis under earthquake lead to massive calculations in high-speed railway seismic design. To address this challenge, this paper introduces a deep learning-based method at enhancing the computational efficiency of the train-bridge coupled (TBC) system’s seismic response prediction. Accordingly, a deep learning framework for predicting stochastic seismic response is established by using a long short-term memory (LSTM) neural network. Meanwhile, a feasible training strategy for the LSTM network is proposed. A traditional TBC theory and the LSTM neural network are used to perform the seismic response analysis as illustrative examples. The comparative analysis demonstrates the LSTM network's remarkable accuracy and efficiency in predicting the TBC system's seismic response. The prediction performance and extrapolation capability of the LSTM network is evaluated to be good enough to meet the requirement of engineering applications. In contrast to conventional methods like TBC theory or finite element analysis, the LSTM network significantly improves computational efficiency. Furthermore, the computed response data and the evolution characteristics of the probability density function are in good agreement with the conventional method. Therefore, the established LSTM network serves as an effective surrogate model for predicting the TBC system’s seismic response.

Advancing Fault Prediction: A Comparative Study between LSTM and Spiking Neural Networks

Predicting system faults is critical to improving productivity, reducing costs, and enforcing safety in industrial processes. Yet, traditional methodologies frequently falter due to the intricate nature of the task. This research presents a novel use of spiking neural networks (SNNs) in anticipating faults in syntactical time series, utilizing the generalized stochastic Petri net (GSPN) model. The inherent ability of SNNs to process both time and space aspects of data positions them as a prime instrument for this endeavor. A comparative evaluation with long short-term memory (LSTM) networks suggests that SNNs offer comparable robustness and performance.

Low-Complexity Conditional Generative Adversarial Network (c-GAN) Based Nonlinear Equalizer for Coherent Data-Center Interconnections

Nonlinear impairments emerging from both fiber optical links and optoelectrical devices are the major bottleneck for enhancing the throughput of photonic data-center interconnection (DCI). Here, we propose a data-driven nonlinear equalizer based on a conditional generative adversarial network (c-GAN) for high-speed and large-capacity photonic DCI. Its performance is experimentally evaluated by transmitting C-band 40-channel 60-GBaud dual polarization-16 quadrature amplitude modulation (DP-16QAM) dense wavelength division multiplexing (DWDM) signals over 20-km standard single-mode fiber (SSMF). Our experimental results verify that, the proposed c-GAN equalizer outperforms both traditional nonlinear equalizers and other data-driven counterparts, in terms of both the bit-error ratio (BER) and computational complexity. Only the proposed c-GAN equalizer can guarantee the BER performance of all 40 channels to reach the threshold of 20% soft decision forward error correction (SD-FEC) at BER=0.027, leading to a net transmission capacity of 16 Tb/s. Meanwhile, the computational complexity of the proposed c-GAN equalizer can be significantly reduced by 31.8%, 80.7%, and 98.8%, respectively, in comparison with Volterra filter equalizer (VFE), multilayer perceptron (MLP), and long short-term memory neural network (LSTM-NN) schemes.

Energy Financial Risk Management in China Using Complex Network Analysis

Effective energy financial risk management is crucial to ensure that China's economic system can remain stable. This article utilizes the quantile vector autoregressive spillover index model, complex networks, and deep learning methods to simultaneously assess both the internal and external energy financial market risks in China. Spillover effects under different market conditions are also examined. The research findings indicate that: (1) Under extreme market conditions, static total spillover values between internal and external markets exceed 70%, while under normal market conditions, they are only around 53% and 13%, respectively; (2) Crude oil and fuel oil as well as energy and stocks are important nodes in both internal and external markets; and (3) The attention-convolutional neural network-long short-term memory model outperforms the second-best performing model, and achieves an improvement of 12.9% and 21.4% in terms of mean absolute error and root mean square error, respectively; inclusion of early warning indicators leads to further improvements of 19.8% and 31.9%, respectively.

The contribution of remote sensing and input feature selection for groundwater level prediction using LSTM neural networks in the Oum Er-Rbia Basin, Morocco

The planning and management of groundwater in the absence of in situ climate data is a delicate task, particularly in arid regions where this resource is crucial for drinking water supplies and irrigation. Here the motivation is to evaluate the role of remote sensing data and Input feature selection method in the Long Short Term Memory (LSTM) neural network for predicting groundwater levels of five wells located in different hydrogeological contexts across the Oum Er-Rbia Basin (OER) in Morocco: irrigated plain, floodplain and low plateau area. As input descriptive variable, four remote sensing variables were used: the Integrated Multi-satellite Retrievals (IMERGE) Global Precipitation Measurement (GPM) precipitation, Moderate resolution Imaging Spectroradiometer (MODIS) normalized difference vegetation index (NDVI), MODIS land surface temperature (LST), and MODIS evapotranspiration. Three LSTM models were developed, rigorously analyzed and compared. The LSTM-XGB-GS model, was optimized using the GridsearchCV method, and uses a single remote sensing variable identified by the input feature selection method XGBoost. Another optimized LSTM model was also constructed, but uses the four remote sensing variables as input (LSTM-GS). Additionally, a standalone LSTM model was established and also incorporating the four variables as inputs. Scatter plots, violin plots, Taylor diagram and three evaluation indices were used to verify the performance of the three models. The overall result showed that the LSTM-XGB-GS model was the most successful, consistently outperforming both the LSTM-GS model and the standalone LSTM model. Its remarkable accuracy is reflected in high R2 values (0.95 to 0.99 during training, 0.72 to 0.99 during testing) and the lowest RMSE values (0.03 to 0.68 m during training, 0.02 to 0.58 m during testing) and MAE values (0.02 to 0.66 m during training, 0.02 to 0.58 m during testing). The LSTM-XGB-GS model reveals how hydrodynamics, climate, and land-use influence groundwater predictions, emphasizing correlations like irrigated land-temperature link and floodplain-NDVI-evapotranspiration interaction for improved predictions. Finally, this study demonstrates the great support that remote sensing data can provide for groundwater prediction using ANN models in conditions where in situ data are lacking.

Prediction of Sea Surface Chlorophyll-a Concentrations Based on Deep Learning and Time-Series Remote Sensing Data

Accurate prediction of future chlorophyll-a (Chl-a) concentrations is of great importance for effective management and early warning of marine ecological systems. However, previous studies primarily focused on chlorophyll-a inversion and reconstruction, while methods for predicting Chl-a concentrations remain limited. To address this issue, we adopted four deep learning approaches, including Convolutional LSTM Network (ConvLSTM), Convolutional Neural Network-Long Short-Term Memory (CNN-LSTM), Eidetic 3D LSTM (E3D-LSTM), and Self-Attention ConvLSTM (SA-ConvLSTM) models, to predict Chl-a over the Yellow Sea and Bohai Sea (YBS) in China. Furthermore, 14 environmental variables obtained from the remote sensing data of Moderate-resolution Imaging Spectroradiometer (MODIS) and ECMWF Reanalysis v5 (ERA5) were utilized to predict the Chl-a concentrations in the study area. The results showed that all four models performed satisfactorily in predicting Chl-a concentrations in the YBS, with SA-ConvLSTM exhibiting a closer approximation to true values. Furthermore, we analyzed the impact of the Self-Attention Memory Module (SAM) on the prediction results. Compared to the ConvLSTM model, the SA-ConvLSTM model integrated with the SAM module better captured subtle large-scale variations within the study area. The SA-ConvLSTM model exhibited the highest prediction accuracy, and the one-month Pearson correlation coefficient reached 0.887. Our study provides an available approach for anticipating Chl-a concentrations over a large area of sea.

Forest Sound Event Detection with Convolutional Recurrent Neural Network-Long Short-Term Memory

Sound event detection tackles an audio environment's complex sound analysis and recognition problem. The process involves localizing and classifying sounds mainly to estimate the start point and end points of the separate sounds and describe each sound. Sound event detection capability relies on the type of sound. Although detecting sequences of distinct temporal sounds is straightforward, the situation becomes complex when the sound is multiple overlapping of much single audio. This situation usually occurs in the forest environment. Therefore, this aim of the paper is to propose a Convolution Recurrent Neural Network-Long Short-Term Memory algorithm to detect an audio signature of intruders in the forest environment. The audio is extracted in the Mel-frequency cepstrum coefficient and fed into the algorithm as an input. Six sound categories are chainsaw, machete, car, hatchet, ambiance, and bike. They were tested using several epochs, batch size, and filter of the layer in the model. The proposed model can achieve an accuracy of 98.52 percent in detecting the audio signature with a suitable parameter selection. In the future, additional types of audio signatures of intruders and further aspects of evaluation can be added to make the algorithm better at detecting intruders in the forest environment.