Long Short-term Memory Autoencoder Research Articles

Regularized Deep LSTM Autoencoder for Phonological Deviation Assessment

In this paper, the Deep Long-short term memory Autoencoder (DLAE), a regularized deep learning model, is proposed for the automatic severity assessment of phonological deviations which are crucial stuttering markers in children. This automatic noninvasive severity assessment plays a paramount role in prevenient diagnosis, progress inference, and post-care for the patients with specific speech disorder. The proposed model is an implementation of a multi-layered Autoencoder in the Encoder–Decoder architecture of the Long-Short Term Memory (LSTM) model with hierarchically appended hidden layers and hidden units. The DLAE has definite advantage over the baseline Autoencoders. During the training phase, the proposed DLAE reconstructs the phonological features in an unsupervised fashion and the latent bottleneck features are extracted from the Encoder. The trained and regularized DLAE model with drop out is then used to predict the severity of the phonological deviation with high precision and classification accuracy compared to the baseline models.

Traffic Flow Prediction Based on Deep Learning in Internet of Vehicles

In Internet of Vehicles (IoV), accurate traffic flow prediction is helpful for analyzing road condition and then timely feedback traffic information to managers as well as travelers. Traditional traffic flow predictions are generally suffering from the performance degradation by over-fitting and manual intervening, which cannot support large-scale and high-dimensional urban road network data. To address this issue, in this paper, a traffic flow prediction framework for urban road network based on deep learning is proposed. Firstly, the feature engineering is introduced to extract the features from a large volume of traffic dataset, with the anomaly nodes eliminated. Next, the big traffic dataset is compressed through the spectral clustering compression scheme. Finally, we designed a hybrid traffic flow prediction scheme based on LSTM (Long Short Term Memory) and Sparse Auto-Encoder (SAE). Experimental results show that our proposed model is superior to other models with an average prediction accuracy approaching 97.7%.

DC Fault Detection and Pulsed Load Monitoring Using Wavelet Transform-Fed LSTM Autoencoders

The extensive deployment of power electronics loads in naval ship power systems indicates the ship electrification is inevitable in future trends. Next-generation warships require high-power density weapons drawing pulse power from the dc grid. A particularly concerning issue is that these pulse loads draw large currents in short periods of time, similar to fault behavior, and maybe indiscernible from a fault. This article proposed a novel machine learning-based algorithm—long short-term memory (LSTM) recurrent neural network (RNN)-based autoencoder (AE) networks to detect dc faults and monitor load conditions applied to naval pulse loads. The novel load monitoring solution presented herein can be applied to any load profile that exhibits repetitive transients during normal operation. The frequency-domain features of the load current are extracted under wavelet transform for the network training to set the network weights and biases. Once the network training is completed, the LSTM RNN-based AE will produce both signal classification and signal reconstruction of the pulse load based on wavelet features of input current. Any faults should yield large reconstruction errors for protective action. Finally, the method is demonstrated in experimental results.

Spam review detection using LSTM autoencoder: an unsupervised approach

The review of online products or services is becoming a major factor in the user’s purchasing decisions. The popularity and influence of online reviews attract spammers who intend to elevate their products or services by writing positive reviews for them and lowering the business of others by writing negative reviews. Traditionally, the spam review identification task is seen as a two-class classification problem. The classification approach requires a labelled dataset to train a model for the environment it is working on. The unavailability of the labelled dataset is a major limitation in the classification approach. To overcome the problem of the labelled dataset, we propose an unsupervised learning model combining long short-term memory (LSTM) networks and autoencoder (LSTM-autoencoder) to distinguish spam reviews from other real reviews. The said model is trained to learn the patterns of real review from the review’s textual details without any label. The experimental results show that our model is able to separate the real and spam review with good accuracy.

Lower Limb Kinematics Trajectory Prediction Using Long Short-Term Memory Neural Networks.

This study determined whether the kinematics of lower limb trajectories during walking could be extrapolated using long short-term memory (LSTM) neural networks. It was hypothesised that LSTM auto encoders could reliably forecast multiple time-step trajectories of the lower limb kinematics, specifically linear acceleration (LA) and angular velocity (AV). Using 3D motion capture, lower limb position–time coordinates were sampled (100 Hz) from six male participants (age 22 ± 2 years, height 1.77 ± 0.02 m, body mass 82 ± 4 kg) who walked for 10 min at 5 km/h on a 0% gradient motor-driven treadmill. These data were fed into an LSTM model with a sliding window of four kinematic variables with 25 samples or time steps: LA and AV for thigh and shank. The LSTM was tested to forecast five samples (i.e., time steps) of the four kinematic input variables. To attain generalisation, the model was trained on a dataset of 2,665 strides from five participants and evaluated on a test set of 1 stride from a sixth participant. The LSTM model learned the lower limb kinematic trajectories using the training samples and tested for generalisation across participants. The forecasting horizon suggested higher model reliability in predicting earlier future trajectories. The mean absolute error (MAE) was evaluated on each variable across the single tested stride, and for the five-sample forecast, it obtained 0.047 m/s2 thigh LA, 0.047 m/s2 shank LA, 0.028 deg/s thigh AV and 0.024 deg/s shank AV. All predicted trajectories were highly correlated with the measured trajectories, with correlation coefficients greater than 0.98. The motion prediction model may have a wide range of applications, such as mitigating the risk of falls or balance loss and improving the human–machine interface for wearable assistive devices.

Heterogeneous Connection and Process Anomaly Detection of Industrial Robot in Intelligent Factory

Since the concept of industry 4.0 was proposed in 2011, the trend of industry 4.0 has been surging around the world. Intelligent factory is one of the main research points in the industry 4.0 era. In order to improve the intelligent level of the factory, the connection-and-cognition ability has to be established for the factory and its equipment. Connection builds data pipes among equipment and systems while cognition automatically turns the data into knowledge. In an intelligent factory, industrial robot plays a leading role. Hence, the aim of this paper is to synthetically study connection and cognition of industrial robots in intelligent factories. To be specific, open platform communications unified architecture (OPC UA) is applied to establish heterogeneous connection of industrial robots with factory management software. A long short-term memory (LSTM) joint auto encoder method is proposed to establish the unsupervised anomaly detection cognition ability for industrial robot process (e.g. grinding, welding and assembling). In summary, this study puts OPC UA and LSTM auto encoder technology together to study heterogeneous connection and process anomaly detection of industrial robots in intelligent factory. The experimental results showed that the proposed method successfully realized heterogeneous connection of an industrial robot and detected process anomaly from the robot built-in sensors’ data.

Towards Efficient Electricity Forecasting in Residential and Commercial Buildings: A Novel Hybrid CNN with a LSTM-AE based Framework.

Due to industrialization and the rising demand for energy, global energy consumption has been rapidly increasing. Recent studies show that the biggest portion of energy is consumed in residential buildings, i.e., in European Union countries up to 40% of the total energy is consumed by households. Most residential buildings and industrial zones are equipped with smart sensors such as metering electric sensors, that are inadequately utilized for better energy management. In this paper, we develop a hybrid convolutional neural network (CNN) with an long short-term memory autoencoder (LSTM-AE) model for future energy prediction in residential and commercial buildings. The central focus of this research work is to utilize the smart meters’ data for energy forecasting in order to enable appropriate energy management in buildings. We performed extensive research using several deep learning-based forecasting models and proposed an optimal hybrid CNN with the LSTM-AE model. To the best of our knowledge, we are the first to incorporate the aforementioned models under the umbrella of a unified framework with some utility preprocessing. Initially, the CNN model extracts features from the input data, which are then fed to the LSTM-encoder to generate encoded sequences. The encoded sequences are decoded by another following LSTM-decoder to advance it to the final dense layer for energy prediction. The experimental results using different evaluation metrics show that the proposed hybrid model works well. Also, it records the smallest value for mean square error (MSE), mean absolute error (MAE), root mean square error (RMSE) and mean absolute percentage error (MAPE) when compared to other state-of-the-art forecasting methods over the UCI residential building dataset. Furthermore, we conducted experiments on Korean commercial building data and the results indicate that our proposed hybrid model is a worthy contribution to energy forecasting.

Feature Extraction Using an RNN Autoencoder for Skeleton-Based Abnormal Gait Recognition

In skeleton-based abnormal gait recognition, using original skeleton data decreases the recognition performance because they contain noise and irrelevant information. Instead of feeding original skeletal gait data to a recognition model, features extracted from the skeleton data are normally used. However, existing feature extraction methods might include laborious processes and it is hard for them to minimize the irrelevant information while preserving the important information. To solve this problem, an automatic feature extraction method using a recurrent neural network (RNN)-based Autoencoder (AE) is proposed in this paper. We extracted features from skeletal gait data by using two RNN AEs: a long short-term memory (LSTM)-based AE (LSTM AE) and a gated recurrent unit (GRU)-based AE (GRU AE). The features of the RNN AEs are compared to the original skeleton data and other existing features. We evaluated the features by feeding them to various discriminative models (DMs) and comparing the recognition performances. The features extracted by using the RNN AEs are more easily recognized and robust than the original skeleton data and other existing features. In particular, the LSTM AE shows a better performance than the GRU AE. Compared to single DMs fed with the original skeleton directly, hybrid models where the features of the RNN AEs are fed to DMs show a higher recognition accuracy with fewer training epochs and learning parameters. Therefore, the proposed automatic feature extraction method improves the performance of skeleton-based abnormal gait recognition by reducing laborious processes and increasing the recognition accuracy effectively.

Tracking of Flexible Brush Tip on Real Canvas: Silhouette-Based and Deep Ensemble Network-Based Approaches

Digital painting is a process of creating a digital artwork using modern human-computer interaction technologies. One of the core enabling technologies is the real-time tracking of user's strokes, which is generally supplied by a digital tablet with a stylus. While the digital tablet technology provides highly accurate tracking, the drawing should be done with a rigid stylus on a plastic surface. This sometimes destroys the realism of drawing, such as interaction with the digital tablet cannot provide the feedback of subtle texture, friction of the paper/fabric canvas and tension of soft painting brush. This becomes particularly problematic for traditional painting artists who are trained with and prefer real painting brush and paper/fabric canvas. Thus, the aim of this work is to present an alternative solution where the user's strokes can be tracked even when the actual brush and canvas are used. To this end, we proposed two approaches for digitally tracking the tip of flexible bristles of a soft brush, so that the painting can be created digitally on a computer. The first approach captures the silhouette of deforming bristles using a pair of well-aligned infra-red (IR) cameras, which extracts the tip from the silhouette, and reconstructs the 2D position of the tip. The second approach predicts the brush tip position through a deep ensemble network-based approach where the relationship between the brush tip position and brush handle pose are trained with our novel model comprising of Long-Short Term Memory Autoencoder and 1-D Convolutional Neural Network. The trained model is used to predict the brush tip position in realtime. Both approaches extensively evaluated through multiple tests. Furthermore, our model outperforms the state-of-the-art models.

Energy Disaggregation via Deep Temporal Dictionary Learning.

This paper presents a novel nonlinear dictionary learning (DL) model to address the energy disaggregation (ED) problem, i.e., decomposing the electricity signal of a home to its operating devices. First, ED is modeled as a new temporal DL problem where a set of dictionary atoms is learned to capture the most representative temporal features of electricity signals. The sparse codes corresponding to these atoms show the contribution of each device in the total electricity consumption. To learn powerful atoms, a novel deep temporal DL (DTDL) model is proposed that computes complex nonlinear dictionaries in the latent space of a long short-term memory autoencoder (LSTM-AE). While the LSTM-AE captures the deep temporal manifold of electricity signals, the DTDL model finds the most representative atoms inside this manifold. To simultaneously optimize the dictionary and the deep temporal manifold, a new optimization algorithm is proposed that alternates between finding the optimal LSTM-AE and the optimal dictionary. To the best of authors' knowledge, DTDL is the only DL model that understands the deep temporal structures of the data. Experiments on the Reference ED Data Set show an outstanding performance compared with the recent state-of-the-art algorithms in terms of precision, recall, accuracy, and F-score.

An AutoEncoder and LSTM-Based Traffic Flow Prediction Method.

Smart cities can effectively improve the quality of urban life. Intelligent Transportation System (ITS) is an important part of smart cities. The accurate and real-time prediction of traffic flow plays an important role in ITSs. To improve the prediction accuracy, we propose a novel traffic flow prediction method, called AutoEncoder Long Short-Term Memory (AE-LSTM) prediction method. In our method, the AutoEncoder is used to obtain the internal relationship of traffic flow by extracting the characteristics of upstream and downstream traffic flow data. Moreover, the Long Short-Term Memory (LSTM) network utilizes the acquired characteristic data and the historical data to predict complex linear traffic flow data. The experimental results show that the AE-LSTM method had higher prediction accuracy. Specifically, the Mean Relative Error (MRE) of the AE-LSTM was reduced by 0.01 compared with the previous prediction methods. In addition, AE-LSTM method also had good stability. For different stations and different dates, the prediction error and fluctuation of the AE-LSTM method was small. Furthermore, the average MRE of AE-LSTM prediction results was 0.06 for six different days.

Detection of Bleeding Events in Electronic Health Record Notes Using Convolutional Neural Network Models Enhanced With Recurrent Neural Network Autoencoders: Deep Learning Approach.

BackgroundBleeding events are common and critical and may cause significant morbidity and mortality. High incidences of bleeding events are associated with cardiovascular disease in patients on anticoagulant therapy. Prompt and accurate detection of bleeding events is essential to prevent serious consequences. As bleeding events are often described in clinical notes, automatic detection of bleeding events from electronic health record (EHR) notes may improve drug-safety surveillance and pharmacovigilance.ObjectiveWe aimed to develop a natural language processing (NLP) system to automatically classify whether an EHR note sentence contains a bleeding event.MethodsWe expert annotated 878 EHR notes (76,577 sentences and 562,630 word-tokens) to identify bleeding events at the sentence level. This annotated corpus was used to train and validate our NLP systems. We developed an innovative hybrid convolutional neural network (CNN) and long short-term memory (LSTM) autoencoder (HCLA) model that integrates a CNN architecture with a bidirectional LSTM (BiLSTM) autoencoder model to leverage large unlabeled EHR data.ResultsHCLA achieved the best area under the receiver operating characteristic curve (0.957) and F1 score (0.938) to identify whether a sentence contains a bleeding event, thereby surpassing the strong baseline support vector machines and other CNN and autoencoder models.ConclusionsBy incorporating a supervised CNN model and a pretrained unsupervised BiLSTM autoencoder, the HCLA achieved high performance in detecting bleeding events.