Search

Investigating the identity, distribution, and evolution of bird species is important for both biodiversity assessment and environmental conservation. The discrete wavelet transform (DWT) has been widely exploited to extract time–frequency features for acoustic signal analysis. Traditional approaches usually compute statistical measures (e.g., maximum, mean, standard deviation) of the DWT coefficients in each subband independently to yield the feature descriptor, without considering the intersubband correlation. A new acoustic descriptor, called the local wavelet acoustic pattern (LWAP), is proposed to characterize the correlation of the DWT coefficients in different subbands for birdsong recognition. First, we divide a variable-length birdsong segment into a number of fixed-duration texture windows. For each texture window, several LWAP descriptors are extracted. The vector of locally aggregated descriptors (VLAD) is then used to aggregate the set of LWAP descriptors into a single VLAD vector. Finally, principal component analysis (PCA) plus linear discriminant analysis (LDA) are employed to reduce the feature dimensionality for classification purposes. Experiments on two birdsong datasets show that the proposed LWAP descriptor outperforms other local descriptors, including linear predictive coding cepstral coefficients, Mel-frequency cepstral coefficients, perceptual linear prediction cepstral coefficients, chroma features, and prosody features. Furthermore, the proposed LWAP descriptor, followed by VLAD encoding, PCA plus LDA feature extraction, and a simple distance-based classifier, yields promising results that are competitive with those obtained by the state-of-the-art convolutional neural networks.

In this paper, a color image representation approach is proposed for color image retrieval. First, an input color image is represented by a number representative images (called color degree images) consisting of chromatic (including red, green, blue, cyan, magenta, and yellow) and achromatic (including black and white) degree images. For each pixel, the degree to which its color value is similar to each representative color will be computed. For the local window centered at each detected key point, a vector representing the degree values, called local color degree (LCD) descriptor, will be computed to describe the local color property. Meanwhile, the scale-invariant feature transform (SIFT) descriptor is extracted to describe the local texture characteristics. Both LCD descriptor and SIFT descriptor are combined together to form the local descriptor. Vector of locally aggregated descriptors (VLAD) is then employed to aggregate all local descriptors to get a single VLAD vector for each input image. Experimental results on INRIA Holidays and Leeds Butterfly datasets have shown that the proposed descriptor yield higher retrieval accuracy.

In this paper, a novel feature line embedding (FLE) algorithm based on support vector machine (SVM), referred to as SVMFLE, is proposed for dimension reduction (DR) and for improving the performance of the generative adversarial network (GAN) in hyperspectral image (HSI) classification. The GAN has successfully shown high discriminative capability in many applications. However, owing to the traditional linear-based principal component analysis (PCA) the pre-processing step in the GAN cannot effectively obtain nonlinear information; to overcome this problem, feature line embedding based on support vector machine (SVMFLE) was proposed. The proposed SVMFLE DR scheme is implemented through two stages. In the first scatter matrix calculation stage, FLE within-class scatter matrix, FLE between-scatter matrix, and support vector-based FLE between-class scatter matrix are obtained. Then in the second weight determination stage, the training sample dispersion indices versus the weight of SVM-based FLE between-class matrix are calculated to determine the best weight between-scatter matrices and obtain the final transformation matrix. Since the reduced feature space obtained by the SVMFLE scheme is much more representative and discriminative than that obtained using conventional schemes, the performance of the GAN in HSI classification is higher. The effectiveness of the proposed SVMFLE scheme with GAN or nearest neighbor (NN) classifiers was evaluated by comparing them with state-of-the-art methods and using three benchmark datasets. According to the experimental results, the performance of the proposed SVMFLE scheme with GAN or NN classifiers was higher than that of the state-of-the-art schemes in three performance indices. Accuracies of 96.3%, 89.2%, and 87.0% were obtained for the Salinas, Pavia University, and Indian Pines Site datasets, respectively. Similarly, this scheme with the NN classifier also achieves 89.8%, 86.0%, and 76.2% accuracy rates for these three datasets.

In this paper, a visual face feature extraction scheme using image processing techniques for health management systems is proposed. Five visual face features, i.e., face contours, face colors, smile lines, hairlines, and melanocytes, were detected, extracted, stored, and retrieved for healthcare management. These actions were performed using the software services in cloud computing environments. Users can periodically browse the changes of the face shapes and texture features. Chinese medical doctors could also check the extracted visual features and give some suggestions for patients via the Internet.

The focus of content-based image retrieval (CBIR) is to narrow down the gap between low-level image features and high-level semantic concepts. In this paper, a biased discriminant analysis with feature line embedding (FLE-BDA) is proposed for performance enhancement in relevance feedback schemes. Maximizing the margin between relevant and irrelevant samples at local neighborhoods was the aim in this study. In reduced subspace, relevant images and query images can be quite close, while irrelevant samples are far away from relevant samples. The results of four benchmark datasets are given to show the performance of the proposed method.

In this study, an assistant system is developed for pharmacist to improve the dispensing quality by two functions: notification in time and monitoring in real time. During drug dispensation, the system gets the patient’s prescription issued from doctors, and drives the light-emitting diode (LED) for notification. Since some drug titles, shapes, colors, or packages are very similar, pharmacists waste lots of time to find the correct drugs. With LED notification, pharmacists pick up the drugs from the correct cabinets, and save the dispensing time. Second, the system monitors pharmacist actions by the infrared (IR) sensors. An alarm is given if pharmacists pick up the incorrect drugs or lost the drug items, even the correct LEDs are turn on. In addition, a web-based information system is designed for drug dispensing and inventory management. During the dispensation, patient information and drug data are displayed on the screen for notification.

Thanks to advances in wireless communication technologies, the wireless sensor network (WSNs) have been attracting a lot of attention from academic communities and successfully applied to various domains. Along with developments of the Industry 4.0, the WSNs start to play a vital role in the construction of smart factories and realization of intelligent manufacturing. Although, the industrial WSNs (IWSNs) presents great quantity of advantages, there still have some drawbacks to overcome such as challenges of the quality of data for IWSNs. In order to resolve the data missing problems in the context of IWSNs, the Last Observation Carried Forward method is adopted to estimate the missing value and reconstruct the sensing dataset which takes into account the temporal characteristics of sensing data in IWSNs. Through experiments, this method is proved to be an easy and effective measurement for missing value imputation of the large multi-dimensional sensing data achieved by the IWSNs.

Recently, teaching using video clips is a simple and direct method instead of the text-based approaches. On the E-learning platform, students ask a question, and teachers answer the question using video clips especially the derivation of mathematical equations. Teachers write their answers down on an A4-size paper, use an APP to record the whole process, and upload the video clips to the platform. However, reading the sentences or equations from videos is very uncomfortable due to the paper movements when teacher wrote the sentences. In addition, teacher's hand frequently occludes the answers during the recording process. In the study, a video stabilizer with hand removal is proposed for comfortably reading. There are two key components: Video stabilization and sentence identification. Video stabilization estimates the smooth camera paths using SURF -based corner detection, hand part removal, image registration, and frame calibration. In sentence identification, sentences are segmented and extracted using contour detection and the convex hull generation algorithm. After the calibration process, users won't fell dizzy in reading sentences from calibrated and un-shaky videos. In addition, time stamps are marked at the beginning and the end of each sentence in the processed videos. When users click a specified sentence, a stabilized clip with good quality will be played.

Robust face recognition under illumination variations is a critical problem in a face recognition system, particularly for face recognition in the wild. In this paper, a face image preprocessing approach, called spatial expansion local histogram equalization (SELHE), is proposed to enhance face images due to illumination variations. First, a face image is divided into several non-overlapped blocks. Then, local histogram equalization with spatial expansion is proposed to enhance the contrast of each local image block. Local linear regression classification will then be used to recognize the enhanced image blocks. Experiments performed on the Yale B and Yale B extended databases have shown that the proposed approach yields promising recognition accuracy.

Robust face recognition under illumination variations is an important and challenging task in a face recognition system, particularly for face recognition in the wild. In this paper, a face image preprocessing approach, called spatial adaptive shadow compensation (SASC), is proposed to eliminate shadows in the face image due to different lighting directions. First, spatial adaptive histogram equalization (SAHE), which uses face intensity prior model, is proposed to enhance the contrast of each local face region without generating visible noises in smooth face areas. Adaptive shadow compensation (ASC), which performs shadow compensation in each local image block, is then used to produce a wellcompensated face image appropriate for face feature extraction and recognition. Finally, null-space linear discriminant analysis (NLDA) is employed to extract discriminant features from SASC compensated images. Experiments performed on the Yale B, Yale B extended, and CMU PIE face databases have shown that the proposed SASC always yields the best face recognition accuracy. That is, SASC is more robust to face recognition under illumination variations than other shadow compensation approaches.

Personal authentication (PA) on smartphones plays the crucial role in mobile payment. Facial features are the most user-friendly biometric feature because of the build-in camera, when we use smartphones as the payment devices. In this study, a novel authenticated method is proposed for PA by integrating feature line embedding (FLE) transformation and biased discriminant analysis (BDA) by using facial features. Due to the few training samples, the discriminant power is limited for learning. In feature spaces, feature lines are regarded as the feature combination between two training samples and infinitely simulate the possible features of various conditions for training. In PA, only positive samples is used to calculate the within-class scatter, and the between class scatter is also calculated using negative samples by the BDA strategy. Compared with the traditional two-class classification and BDA problems, the FLE integrates with BDA to obtain a better dimension reduction transformation. A support vector machine (SVM) classifier is further trained to determine a query sample is a real or a forged sample.

In this paper, a new 2D cepstral descriptor will be proposed for 3D model retrieval. In the proposed approach, the overlapping subband decomposition will be employed to divide each 2D spectrum into a number of overlapped subbands for further 2D cepstral descriptor extraction. Experiments conducted on the Princeton Shape Benchmark (PSB) database have shown that the proposed 2D cepstral descriptor yields better retrieval accuracy than traditional 2D cepstral descriptors and outperforms other state-of-the-art descriptors in terms of the discounted cumulative gain (DCG) score.

Eigenspace projection methods are widely used for feature extraction from hyperspectral images (HSI) for the classification of land cover. Projection transformation is used to reduce higher dimensional feature vectors to lower dimensional vectors for more accurate classification of land cover types. In this paper, a nearest feature line embedding (NFLE) transformation is proposed for the dimension reduction (DR) of an HSI. The NFL measurement is embedded in the transformation during the discriminant analysis phase, instead of the matching phase. Three factors, including class separability, neighborhood structure preservation, and NFL measurement, are considered simultaneously to determine an effective and discriminating transformation in the eigenspaces for land cover classification. Three state-of-the-art classifiers, the nearest-neighbor, support vector machine, and NFL classifiers, were used to classify the reduced features. The proposed NFLE transformation is compared with different feature extraction approaches and evaluated using two benchmark data sets, the MASTER set at Au-Ku and the AVIRIS set at Northwest Tippecanoe County. The experimental results demonstrate that the NFLE approach is effective for DR in land cover classification in the field of Earth remote sensing.

Chinese medicine is a time-honored and deeply valued medical skill in the Chinese community. Although currently the scientific Chinese medicine powder is prevalent, its varieties are numerous and have similar appearance, color, and odor, resulting in difficulties in being distinguished. Moreover, dosages are still weighed and mixed manually, rather than being automated, therefore, the workload for each Chinese medicine pharmacist is heavy, leading to it being easy for the wrong medicine to be administered due to human negligence. An innovative shelf-based automatic powder dispensing machine is therefore presented in this paper in order to promote the quality and efficiency of powder dispensing. This machine consists of linear guide mechanisms, weighing mechanisms, retractable mechanisms, gallipot base discharge mechanisms, and a lifting platform mechanism, and offers several innovative or important features that are beneficial to Chinese medicine. It can automatically and accurately dispense different powders in the correct weights, reducing the possibility of human error. Each gallipot is equipped with one low-cost gallipot base, which can be controlled by the weighing mechanism to discharge powder with the desired weight. Powder cross-contamination among different gallipots can also be avoided. Stepping motors are used in the presented machine for easy control, low cost, and durability. A parabolic type of S mode speed control for stepping motor is implemented, providing faster and more stable moving speed and accuracy than the traditional trapezoidal mode. Hardware and software are integrated and tested. The error of the weighed powder is reduced to +/&minus, 0.2 g. The dispensing process is very efficient, the moving speed of the weighing mechanism reaches 23.89 cm/s, and the discharge speed is up to 0.22 g/s. The system is indeed able to improve the accuracy and efficiency of powder dispensing for Chinese medicine.

The Chenyulan watershed, located in the central mountain area of Taiwan, has been suffering from earthquakes, typhoons, and heavy rainfalls in recent decades. These sequential natural disturbances have a cumulative impact on the watershed, leading to more fragile and fragmented land cover and loss of capacity of soil water conservation. In this study, the Soil and Water Assessment Tool (SWAT) and a landscape metrics tool (FRAGSTATS) were used to assess the direct impact (e.g., by annual rainfall) and indirect impact (e.g., by landscape configuration and composition) of natural disturbances on the ecohydrological processes of the Chenyulan watershed. Six SPOT satellite images from 2008 to 2013 were analyzed by using the nearest feature line embedding (NFLE) approach and reclassified into six land cover types: forest, cultivated land, grassland, river, landslide, and built-up. Forest was found to have the largest patch size, indicating that it is more resilient to disturbances, while agricultural land tended to expand from the river side toward the hill. Two land cover change scenarios were compared in the SWAT model. The results showed that there was no significant difference in simulated streamflow during 2004&ndash, 2015 and sediment loading during 2004&ndash, 2009, however, the model performed better for sediment loading during 2010&ndash, 2015 with dynamic land cover change (coefficient of determination (R2) = 0.66, Nash-Sutcliffe efficiency coefficient (NSE) = 0.62, percent bias (PBIAS) = 10.5%, root mean square error observation standard deviation ratio (RSR) = 0.62) than with constant land cover (R2 = 0.61, NSE = 0.54, PBIAS = &minus, 17.3%, RSR = 0.68), indicating that long-term land cover change should be considered in hydrologic modeling. Changes in landslides during 2008&ndash, 2013 were found to significantly affect ecohydrological processes, especially after 2011. In general, annual precipitation plays a dominant role, and landscape composition had by far the strongest influence on water yield and sediment yield compared to landscape configuration. The results can be useful for understanding the effects of land cover change on ecohydrological processes in the Chenyulan watershed and the potential impact of ecohydrological changes on the environment and public health.