Search

The present work analyzes the application of deep learning in the context of digital twins (DTs) to promote the development of smart cities. According to the theoretical basis of DTs and the smart city construction, the five-dimensional DTs model is discussed to propose the conceptual framework of the DTs city. Then, edge computing technology is introduced to build an intelligent traffic perception system based on edge computing combined with DTs. Moreover, to improve the traffic scene recognition accuracy, the Single Shot MultiBox Detector (SSD) algorithm is optimized by the residual network, form the SSD-ResNet50 algorithm, and the DarkNet-53 is also improved. Finally, experiments are conducted to verify the effects of the improved algorithms and the data enhancement method. The experimental results indicate that the SSD-ResNet50 and the improved DarkNet-53 algorithm show fast training speed, high recognition accuracy, and favorable training effect. Compared with the original algorithms, the recognition time of the SSD-ResNet50 algorithm and the improved DarkNet-53 algorithm is reduced by 6.37ms and 4.25ms, respectively. The data enhancement method used in the present work is not only suitable for the algorithms reported here, but also has a good influence on other deep learning algorithms. Moreover, SSD-ResNet50 and improved DarkNet-53 algorithms have significant applicable advantages in the research of traffic sign target recognition. The rigorous research with appropriate methods and comprehensive results can offer effective reference for subsequent research on DTs cities.

Convolutional neural networks are widely used in artificial intelligence and Internet of Things area. As the scale of convolutional neural network expands, more and more processing units are provided for it. The systems are easy prone to error, and any computing problems in any layer of the network will lead to wrong output results. Traditional multimode redundancy methods make the systems more complex, and increase power consumption. This paper proposes an autonomous error-tolerant architecture for convolutional neural networks. Taking the LeNet-5 as an example, the network layers of CNN are mapped on the AET architecture, an error-tolerant synapse is designed to discover the errors, an active evolution scheme is designed to handle unrecoverable errors and implement network reconfiguration. This design is implemented on FPGA, and the experimental results show that this architecture can realize effective error tolerance for convolutional neural network and has fast error recovery ability under the premise of ensuring the same recognition accuracy. © 2020 IEEE., QC 20210915

Biometric systems are increasingly replacing traditional password- and token-based authentication systems. Security and recognition accuracy are the two most important aspects to consider in designing a biometric system. In this paper, a comprehensive review is presented to shed light on the latest developments in the study of fingerprint-based biometrics covering these two aspects with a view to improving system security and recognition accuracy. Based on a thorough analysis and discussion, limitations of existing research work are outlined and suggestions for future work are provided. It is shown in the paper that researchers continue to face challenges in tackling the two most critical attacks to biometric systems, namely, attacks to the user interface and template databases. How to design proper countermeasures to thwart these attacks, thereby providing strong security and yet at the same time maintaining high recognition accuracy, is a hot research topic currently, as well as in the foreseeable future. Moreover, recognition accuracy under non-ideal conditions is more likely to be unsatisfactory and thus needs particular attention in biometric system design. Related challenges and current research trends are also outlined in this paper.

Biometric systems are increasingly replacing traditional password- and token-based authentication systems. Security and recognition accuracy are the two most important aspects to consider in designing a biometric system. In this paper, a comprehensive review is presented to shed light on the latest developments in the study of fingerprint-based biometrics covering these two aspects with a view to improving system security and recognition accuracy. Based on a thorough analysis and discussion, limitations of existing research work are outlined and suggestions for future work are provided. It is shown in the paper that researchers continue to face challenges in tackling the two most critical attacks to biometric systems, namely, attacks to the user interface and template databases. How to design proper countermeasures to thwart these attacks, thereby providing strong security and yet at the same time maintaining high recognition accuracy, is a hot research topic currently, as well as in the foreseeable future. Moreover, recognition accuracy under non-ideal conditions is more likely to be unsatisfactory and thus needs particular attention in biometric system design. Related challenges and current research trends are also outlined in this paper.

In this paper, two approaches to building extraction from satellite imagery and height data obtained from stereo images or LIDAR are compared. The first approach consists of detecting high-rise objects in a digital surface model and then improving recognition accuracy using segmentation of spectral information. The second approach uses the U-Net convolutional neural network, which showed the best results for the extraction of objects from aerospace images on a number of large datasets. Extensive experiments were carried out to evaluate the dependence of the quality of U-Net-based building extraction on the different data types (including high-resolution satellite images and digital surface model data). Building extraction quality of the trained network was also evaluated on satellite images with different spatial resolutions. © 2018 CEUR-WS. All rights reserved.

The performance of automatic speech recognition based on coded-decoded speech heavily depends on the quality of the transmitted signals, determined by channel impairments. This paper examines relationships between speech recognition performance and measurements of speech quality and intelligibility over transmission channels. Different to previous studies, the effects of super-wideband transmissions are analyzed and compared to those of wideband and narrowband channels. Furthermore, intelligibility scores, gathered by conducting a listening test based on logatomes, are also considered for the prediction of automatic speech recognition results. The modern instrumental measurement techniques POLQA and POLQA-based intelligibility have been respectively applied to estimate the quality and the intelligibility of transmitted speech. Based on our results, polynomial models are proposed that permit the prediction of speech recognition accuracy from the subjective and instrumental measures, involving a number of channel distortions in the three bandwidths. This approach can save the costs of performing automatic speech recognition experiments and can be seen as a first step towards a useful tool for communication channel designers. Copyright © 2017 ISCA. Amazon Alexa; Apple; DiDi; et al.; Furhat Robotics; Microsoft

There is a great interest in dimensionality reduction techniques for tackling the problem of high-dimensional pattern classification. This paper addresses the topic of supervised learning of a linear dimension reduction mapping suitable for classification problems. The proposed optimization procedure is based on minimizing an estimation of the nearest neighbor classifier error probability, and it learns a linear projection and a small set of prototypes that support the class boundaries. The learned classifier has the property of being very computationally efficient, making the classification much faster than state-of-the-art classifiers, such as SVMs, while having competitive recognition accuracy. The approach has been assessed through a series of experiments, showing a uniformly good behavior, and competitive compared with some recently proposed supervised dimensionality reduction techniques. © 2010 Elsevier B.V. All rights reserved.

There is a great interest in dimensionality reduction techniques for tackling the problem of high-dimensional pattern classification. This paper addresses the topic of supervised learning of a linear dimension reduction mapping suitable for classification problems. The proposed optimization procedure is based on minimizing an estimation of the nearest neighbor classifier error probability, and it learns a linear projection and a small set of prototypes that support the class boundaries. The learned classifier has the property of being very computationally efficient, making the classification much faster than state-of-the-art classifiers, such as SVMs, while having competitive recognition accuracy. The approach has been assessed through a series of experiments, showing a uniformly good behavior, and competitive compared with some recently proposed supervised dimensionality reduction techniques. © 2010 Elsevier B.V. All rights reserved.

Recently, we proposed an improvement to the eigenvoice (EV) speaker adaptation called kernel eigenvoice (KEV) speaker adaptation. In KEV adaptation, eigenvoices are computed using kernel PCA, and a new speaker’s adapted model is implicitly computed in the kernel-induced feature space. Due to many online kernel evaluations, both adaptation and subsequent recognition of KEV adaptation are slower than EV adaptation. In this paper, we eliminate all online kernel computations by finding an approximate pre-image of the implicit adapted model found by KEV adaptation. Furthermore, the two steps of finding the implicit adapted model and its approximate pre-image are integrated by embedding the kernel PCA procedure in our new embedded kernel eigenvoice (eKEV) speaker adaptation method. When tested in an TIDIGITS task with less than 10s of adaptation speech, eKEV adaptation obtained a speedup of 6–14 times in adaptation and 136 times in recognition over KEV adaptation with 12–13% relative improvement in recognition accuracy.

Recently, we proposed an improvement to the eigenvoice (EV) speaker adaptation called kernel eigenvoice (KEV) speaker adaptation. In KEV adaptation, eigenvoices are computed using kernel PCA, and a new speaker’s adapted model is implicitly computed in the kernel-induced feature space. Due to many online kernel evaluations, both adaptation and subsequent recognition of KEV adaptation are slower than EV adaptation. In this paper, we eliminate all online kernel computations by finding an approximate pre-image of the implicit adapted model found by KEV adaptation. Furthermore, the two steps of finding the implicit adapted model and its approximate pre-image are integrated by embedding the kernel PCA procedure in our new embedded kernel eigenvoice (eKEV) speaker adaptation method. When tested in an TIDIGITS task with less than 10s of adaptation speech, eKEV adaptation obtained a speedup of 6–14 times in adaptation and 136 times in recognition over KEV adaptation with 12–13% relative improvement in recognition accuracy.

Recently, we proposed an improvement to the eigenvoice (EV) speaker adaptation called kernel eigenvoice (KEV) speaker adaptation. In KEV adaptation, eigenvoices are computed using kernel PCA, and a new speaker’s adapted model is implicitly computed in the kernel-induced feature space. Due to many online kernel evaluations, both adaptation and subsequent recognition of KEV adaptation are slower than EV adaptation. In this paper, we eliminate all online kernel computations by finding an approximate pre-image of the implicit adapted model found by KEV adaptation. Furthermore, the two steps of finding the implicit adapted model and its approximate pre-image are integrated by embedding the kernel PCA procedure in our new embedded kernel eigenvoice (eKEV) speaker adaptation method. When tested in an TIDIGITS task with less than 10s of adaptation speech, eKEV adaptation obtained a speedup of 6–14 times in adaptation and 136 times in recognition over KEV adaptation with 12–13% relative improvement in recognition accuracy.

Nowadays, almost all speaker-independent (SI) speech recognition systems use CDHMM with multivariate mixture Gaussian as observation density to cover speaker variabilities. It has been shown that given sufficient training data, the more mixtures are used in the HMM observation density, the better the system's perform. However, acoustic HMM with more Gaussian densities is more complex and slows down recognition speed. Another efficient way to handle speaker variation is to use speaker adaptation (SA). Yet, even though speaker adaptation of full multivariate mixture Gaussian densities can increase recognition accuracy, it does not improve recognition speed. In this paper, we introduce a principal mixture speaker adaptation method which reduces HMM complexity by choosing only the principle mixtures corresponding to a particular speaker's characteristics. We show that our method both improves recognition accuracy by 31.8% when compared to SI models, and reduces recognition speed by 30%, when compared to full mixture SA models.

Nowadays, almost all speaker-independent (SI) speech recognition systems use CDHMM with multivariate mixture Gaussian as observation density to cover speaker variabilities. It has been shown that given sufficient training data, the more mixtures are used in the HMM observation density, the better the system's perform. However, acoustic HMM with more Gaussian densities is more complex and slows down recognition speed. Another efficient way to handle speaker variation is to use speaker adaptation (SA). Yet, even though speaker adaptation of full multivariate mixture Gaussian densities can increase recognition accuracy, it does not improve recognition speed. In this paper, we introduce a principal mixture speaker adaptation method which reduces HMM complexity by choosing only the principle mixtures corresponding to a particular speaker's characteristics. We show that our method both improves recognition accuracy by 31.8% when compared to SI models, and reduces recognition speed by 30%, when compared to full mixture SA models.

Nowadays, almost all speaker-independent (SI) speech recognition systems use CDHMM with multivariate mixture Gaussian as observation density to cover speaker variabilities. It has been shown that given sufficient training data, the more mixtures are used in the HMM observation density, the better the system's perform. However, acoustic HMM with more Gaussian densities is more complex and slows down recognition speed. Another efficient way to handle speaker variation is to use speaker adaptation (SA). Yet, even though speaker adaptation of full multivariate mixture Gaussian densities can increase recognition accuracy, it does not improve recognition speed. In this paper, we introduce a principal mixture speaker adaptation method which reduces HMM complexity by choosing only the principle mixtures corresponding to a particular speaker's characteristics. We show that our method both improves recognition accuracy by 31.8% when compared to SI models, and reduces recognition speed by 30%, when compared to full mixture SA models.

Yang, W., Wang, S., Hu, J., Zheng, G., & Valli, C. (2019). Security and Accuracy of Fingerprint-Based Biometrics: A Review. Symmetry, 11(2), 141. Available here.

The own-race bias (ORB) is a well-documented phenomenon whereby people are more accurate at recognising faces from their own-race group compared to faces from different race groups (Meissner & Brigham, 2001). While empirical studies have reported the robustness of the ORB across racial groups, the finding that the ORB is negatively related to exposure to other-race individuals suggests that may not necessarily apply to individuals who were born and have lived in a multiracial society. This thesis fills a gap in the ORB literature by examining the role of multiracial immersion experience as well as potential perceptual/attentional mechanisms in modulating the ORB by employing a series of face memory and face perceptual tasks on different racial groups (i.e. Malaysian-Malay, Malaysian-Chinese, Malaysian-Indian, African and Western-Caucasian individuals). In Experiment 1, I demonstrated that the ORB in face memory was pronounced not only in Western-Caucasian participants, but also in Malaysian participants (Malay, Chinese, and Indian) who were raised in a racially diverse environment, implying that growing up in a profoundly multiracial society does not eliminate the ORB. In Experiment 2, I found significant correlations between memory performance for learned target faces and facial distinctiveness; most importantly, own-race faces benefited more from this distinctiveness effect than other-race faces across the four race groups. These findings fit well with Valentine’s (1991) face-space model that dimensions of face space are optimised for own-race faces, and the ORB is likely to be attributable to the inefficiency in the use of norm-based coding for other-race faces. Using a procedure identical to Experiment 1, I observed a significantly greater increment in recognition performance for other-race faces than for own-race faces when the external features (e.g., facial contour and hairline) were presented along with the internal features in Experiment 3. This finding sup