Search

Recent advances in wearable systems have made inertial sensors, such as accelerometers and gyroscopes, compact, lightweight, multimodal, low-cost, and highly accurate. Wearable inertial sensor-based multimodal human activity recognition (HAR) methods utilize the rich sensing data from embedded multimodal sensors to infer human activities. However, existing HAR approaches either rely on domain knowledge or fail to address the time-frequency dependencies of multimodal sensor signals. In this paper, we propose a novel method called deep wavelet convolutional neural networks (DWCNN) designed to learn features from the time-frequency domain and improve accuracy for multimodal HAR. DWCNN introduces a framework that combines continuous wavelet transforms (CWT) with enhanced deep convolutional neural networks (DCNN) to capture the dependencies of sensing signals in the time-frequency domain, thereby enhancing the feature representation ability for multiple wearable inertial sensor-based HAR tasks. Within the CWT, we further propose an algorithm to estimate the wavelet scale parameter. This helps enhance the performance of CWT when computing the time-frequency representation of the input signals. The output of the CWT then serves as input for the proposed DCNN, which consists of residual blocks for extracting features from different modalities and attention blocks for fusing these features of multimodal signals. We conducted extensive experiments on five benchmark HAR datasets: WISDM, UCI-HAR, Heterogeneous, PAMAP2, and UniMiB SHAR. The experimental results demonstrate the superior performance of the proposed model over existing competitors. [ABSTRACT FROM AUTHOR]

Lyrics-to-melody generation is an interesting and challenging topic in AI music research field. Due to the difficulty of learning the correlations between lyrics and melody, previous methods suffer from low generation quality and lack of controllability. Controllability of generative models enables human interaction with models to generate desired contents, which is especially important in music generation tasks towards human-centered AI that can facilitate musicians in creative activities. To address these issues, we propose a controllable lyrics-to-melody generation network, ConL2M, which is able to generate realistic melodies from lyrics in user-desired musical style. Our work contains three main novelties: (1) to model the dependencies of music attributes cross multiple sequences, inter-branch memory fusion (Memofu) is proposed to enable information flow between multi-branch stacked LSTM architecture; (2) reference style embedding (RSE) is proposed to improve the quality of generation as well as control the musical style of generated melodies; (3) sequence-level statistical loss (SeqLoss) is proposed to help the model learn sequence-level features of melodies given lyrics. Verified by evaluation metrics for music quality and controllability, initial study of controllable lyrics-to-melody generation shows better generation quality and the feasibility of interacting with users to generate the melodies in desired musical styles when given lyrics. [ABSTRACT FROM AUTHOR]

Bridge weigh-in-motion (BWIM) serves as a method to obtain the weight of passing vehicles from bridge responses. Most BWIM systems proposed so far rely on the measurement of bridge global vibration data, usually strain, to determine the vehicle load. However, because the bridge's global response is sensitive to all vehicles on the bridge, the global vibration-based BWIM techniques usually suffer from inaccuracy in the case where multiple vehicles are present on the bridge. In this paper, a data-driven approach is proposed to extract the passing vehicle's weight and driving speed from vertical acceleration at the bridge joint. As a type of local vibration, the impulse acceleration responses at a bridge joint can be recorded only during a short period when a vehicle is passing over the joint and are thus not sensitive to vehicles at other locations of the bridge. A field test is conducted at a bridge to prepare labeled training data for the use of a convolutional neural network. One accelerometer is installed on the bridge joint to record impulse acceleration, while the vehicle's weight and driving speed are obtained from a WIM station and a camera near the bridge, respectively. A network that detects the vehicle's passage as well as its passing lane is first proposed, followed by a 1-D convolutional neural network that uses the raw data of acceleration as the input to predict the vehicle's gross weight and driving speed. A comparison is made between the 1-D network and an updated 2-D network that uses the wavelet coefficients as the input matrix. The latter one shows better performance, indicating that it is important to choose the proper input data for the network to be trained. A transfer learning technique is used to test the feasibility of the proposed method. Results show that the proposed method can be extended with limited data to bridges other than the bridge where the network is trained. [ABSTRACT FROM AUTHOR]

Passwordless authentication is a trending theme in cyber security, while biometrics gradually replace knowledge-based schemes. However, Personal Identification Numbers, passcodes, and graphical passwords are still considered as the primary means for authentication. Passwords must be memorable to be usable; therefore, users tend to choose easy to guess secrets, compromising security. The Android Pattern Unlock is a popular graphical password scheme that can be easily attacked by exploiting human behavioristic traits. Despite its vulnerabilities, the popularity of the scheme has led researchers to propose adjustments and variations that enhance security but maintain its familiar user interface. Nevertheless, prior work demonstrated that improving security while preserving usability remains frequently a hard task. In this paper we propose a novel graphical password scheme built on the foundations of the well-accepted Android Pattern Unlock method, which is usable, inclusive, universal, and robust against shoulder surfing and (basically) smudge attacks. Our scheme, named Bu-Dash, features a dynamic user interface that mutates every time a user swipes the screen. Our pilot studies illustrate that Bu-Dash attracts positive user acceptance rates, it is secure, and maintains high usability levels. We define complexity metrics that can be used to further diversify user input, and we conduct complexity and security assessments. [ABSTRACT FROM AUTHOR]

One of the most popular recommender system techniques is collaborative filtering (CF). Nowadays, many researchers apply a neural network with CF, but few focus on the neighbors' concept of CF. This work needs to consider two major issues: the similarity levels between the neighbors and the target user and the user's rating pattern conversion. Because different neighbors have a different influence on the target user and different users usually have a different rating pattern, the ratings directly utilized by the neighbor's preference pattern may be incorrect for the target user. Under two main issues, we try to accomplish three main ideas of CF's prediction: the similarity between users' levels, the neighbor's rating, and the rating conversion. Thus, we propose three main modules, the rating conversion module, the similarity module, and the prediction module, to solve the two issues mentioned above. The rating conversion module projects the neighbor's rating into the target user's aspect. The similarity module uses the users' attentions to compute similarity levels between users. Finally, these similarity levels and the converted ratings are integrated to perform the prediction. The proposed method is compared with the current CF with friends and latent factor model using two types of datasets: real-world and synthetic datasets. We evaluate N neighbors and all neighbors on real-world datasets to prove the number of neighbor is important. Moreover, the performance of the rating conversion module is also evaluated. The proposed method simulates the full rating datasets and the partial rating dataset to compare the effectiveness of using different types of distribution and dataset size. The experimental results demonstrate that the proposed method effectively outperformed the baselines using ranking evaluation and prediction accuracy on real-world and synthetic datasets. Besides, The effectiveness of using different the number of neighbors depends on the quality of neighbors. [ABSTRACT FROM AUTHOR]

Bridge weigh-in-motion (BWIM) is a technique for detecting overloaded vehicles crossing a bridge without requiring them to stop. It may also be useful for monitoring the structural health of the bridge itself. To achieve accurate weighing of each vehicle, its properties, such as speed, locus, and wheel positions, should be estimated in advance. Conventionally, such information has been obtained via additional sensors such as cameras or via peak-signal detection, using multiple sensors installed across the bridge. This may require substantial computational resources or expensive synchronization between sensors, and the complexity of the overall BWIM system may lead to frequent breakdowns. In this paper, we propose a single-sensor-based BWIM system that utilizes a deep neural network. First, a vehicle’s properties are obtained via feature extraction from the bridge strain response, as sampled by a single strain sensor. BWIM is then performed, using the same response data. The model parameters for vehicle detection are optimized automatically by consulting a surveillance camera while obtaining ground-truth data for a large number of vehicles crossing the bridge. After the model is optimized for the target bridge, the camera may be removed. Our proposal paves the way toward low-cost, compact, and single-sensor BWIM systems. [ABSTRACT FROM AUTHOR]

Purpose This paper aims to propose a trust-based ant-colony recommender system. It achieves high accuracy and coverage by integrating the importance level of friends. This paper has two main contributions, namely, selecting higher-quality raters and improving the prediction step. From these two contributions, the proposed trust-based ant-colony recommender system could provide more accurate and wider-coverage prediction than existing systems.Design/methodology/approach To obtain higher-quality raters, the data set was preprocessed, and then, trust values were calculated. The depth of search was increased to obtain higher coverage levels. This work also focuses on the importance level of friends in the system. Because the levels of influence on the active user of all friends are not equal, the importance level of friends is integrated into the system by transposing rater's rating to the active user's perspective and then assigning a weight to each rater.Findings The experimental evaluation clearly demonstrates that the proposed method achieves better results in terms of both accuracy and coverage than existing trust-based recommender systems. It was found that integrating the importance level of friends into the system, which transposes ratings and assigns weight to each user, can increase accuracy and coverage.Originality/value Existing trust-based ant-colony recommender systems do not consider the importance level of friends in the prediction step. Most of them only focus on finding raters and then using the rater's real ratings in the prediction step. A new method is proposed that integrates the importance level of friends into the system by transposing a rater's rating to match the active user's perspective and assigning a weight for each rater. The experimental evaluation demonstrates that the proposed method achieves better accuracy and coverage than existing systems. [ABSTRACT FROM AUTHOR]

This paper considers the problem of identifying all locations of subtrees in a large tree or in a large collection of trees that are similar to a specified pattern tree, where all trees are assumed to be rooted and node-labeled. The tree edit distance is a widely-used measure of tree (dis-)similarity, but is NP-hard to compute for unordered trees. To cope with this issue, we propose a new similarity measure which extends the concept of unordered tree inclusion by taking the costs of insertion and substitution operations on the pattern tree into account, and present an algorithm for computing it. Our algorithm has the same time complexity as the original one for unordered tree inclusion, i.e., it runs in $O(T_1T_2)$ time, where $T_1$ and $T_2$ denote the pattern tree and the text tree, respectively, when the maximum outdegree of $T_1$ is bounded by a constant. Our experimental evaluation using synthetic and real datasets confirms that the proposed algorithm is fast and scalable and very useful for bibliographic matching, which is a typical entity resolution problem for tree-structured data. Furthermore, we extend our algorithm to also allow a constant number of deletion operations on $T_1$ while still running in $O(T_1T_2)$ time. [ABSTRACT FROM AUTHOR]

Traffic congestion occurs frequently in urban settings, and is not always caused by traffic incidents. In this paper, we propose a simple method for detecting traffic incidents from probe-car data by identifying unusual events that distinguish incidents from spontaneous congestion. First, we introduce a traffic state model based on a probabilistic topic model to describe the traffic states for a variety of roads. Formulas for estimating the model parameters are derived, so that the model of usual traffic can be learned using an expectation–maximization algorithm. Next, we propose several divergence functions to evaluate differences between the current and usual traffic states and streaming algorithms that detect high-divergence segments in real time. We conducted an experiment with data collected for the entire Shuto Expressway system in Tokyo during 2010 and 2011. The results showed that our method discriminates successfully between anomalous car trajectories and the more usual, slowly moving traffic patterns. [ABSTRACT FROM AUTHOR]