Self-supervised pre-training methods based on contrastive learning or regression tasks can utilize more unlabeled data to improve the performance of automatic speech recognition (ASR). However, the robustness impact of combining the two pre-training tasks and constructing different negative samples for contrastive learning still remains unclear. In this paper, we propose a noise-robust data2vec for self-supervised speech representation learning by jointly optimizing the contrastive learning and regression tasks in the pre-training stage. Furthermore, we present two improved methods to facilitate contrastive learning. More specifically, we first propose to construct patch-based non-semantic negative samples to boost the noise robustness of the pre-training model, which is achieved by dividing the features into patches at different sizes (i.e., so-called negative samples). Second, by analyzing the distribution of positive and negative samples, we propose to remove the easily distinguishable negative samples to improve the discriminative capacity for pre-training models. Experimental results on the CHiME-4 dataset show that our method is able to improve the performance of the pre-trained model in noisy scenarios. We find that joint training of the contrastive learning and regression tasks can avoid the model collapse to some extent compared to only training the regression task.

The disease-free growth of a plant is highly influential for both environment and human life, as numerous microorganisms/viruses/fungus may affect the growth and agricultural production of a plant. Early detection and treatment thus becomes necessary and must be treated on time. The existing vision techniques either involve image segmentation or feature classification/regression applied over aerial images. This results in an increase in time and cost consumption due to various challenges, such as generalization ability and learning cost. Therefore, a feature-based disease detection approach with minimal learning time and generalization ability could be fairly befitting such as an extreme learning machine (ELM). In this letter, we demonstrate an algorithm, L1-ELM, after employing Kuan filtering for preprocessing and different feature computations. At the evaluation stage, the experimentation performed over benchmark plant datasets confirms that L1-ELM outperforms all existing one-class classification algorithms, preserving optimal learning and better generalization.

Energy efficiency is one of the important parameters in cloud computing which is managed by the data centres. Data centres are computer warehouses that are responsible for storing large volumes of data to deal with the daily transaction handling needs of different productions. Effective scheduling for the execution of the request on machines is still a problem. In addition, the power consumption, as well as management of the node clusters is also a problematic situation when the CPU utilisation increases up to the limit. In this paper, efficient minimum execution and completion time scheduling are accomplished by using a machine learning approach for effectual CPU usage and service level agreement fulfilment in data centres, considered in terms of average accuracy which will reduce costs for the maintenance of the data centres in real-time scenarios. The simulation of the proposed work is achieved and the performance is evaluated in terms of power consumption and CPU usage. The proposed research utilises the neural network and linear regression analysis to perform the classification and compares the performance for the efficient CPU usage.

The utilisation of communication network models is growing day by day in a drastic way around the world. In order to communicate the data round the devices, the utilisation or the establishment of communication networks is mandatory. In this paper, a forked communication network model with the base of a queuing model and the equations developed for a better understanding of the model. The results calculated with the help of MATLAB and MathCAD Software. The results observed in detail in the results section. To better understand the performance of the current forked model of communication network, the performance metrics of the current network model had compared with the performance metrics of the previous two three-node communication networks. The comparison done with the arrival process of packets to the network in one model follows the Duane process. In other model follows the Poisson process, whereas in the current model follows the Homogeneous Poisson process of arrivals. The comparison had given in detail in the comparison section.

For implanted neural-sensing devices, one of the remaining challenges is to transmit stable power/data (P/D) transmission for high spatiotemporal resolution neural data. This paper presents a miniaturized implantable 128-channel wireless neural-sensing microsystem using TSV-embedded dissolvable μ-needle array, a flexible interposer and 4 dies by 2.5D/3D TSV heterogeneous SiP technology. The 4 dies are 2 neural-signal acquisition ICs implemented by 90nm CMOS, 1 neural-signal processor by 40nm CMOS and 1 wireless P/D transmission circuitry by 0.18μm CMOS. Thus, the proposed wireless microsystem realizes 128-channel neural-signal sensing within the area of 5mm × 5mm, neural feature extraction and wireless P/D transmission using an on-interposer inductor. The overall average power of the circuits in this microsystem is only 9.85mW.

Highly integrated neural sensing microsystems are crucial to capture accurate signals for brain function investigations. In this paper, a 256-channel/25 mm2 neural sensing microsystem is presented based on through-silicon-via (TSV) 2.5D integration. This microsystem composes of dissolvable μ-needles, TSV-embedded μ-probes, 256-channel neural amplifiers, 11-bit area-power-efficient SAR ADCs and serializers. Based on the dissolvable μ-needles and TSV 2.5D integration, this microsystem can detect 256 ECoG/LFP signals within the small area of 5mm × 5mm. Additionally, the neural amplifier realizes 57.8dB gain with only 9.8μW for each channel, and the 9.7-bit ENOB of the SAR ADC at 32kS/s can be achieved with 0.42μW and 0.036 mm2. The overall power of this microsystem is only 3.79mW for 256-channel neural sensing.

Nowadays, with the explosive growth of the network technologies many new applications and services have been developed on Internet. World Wide Web can provide these services provided without the limitation of time and location. Obviously, the number of user is dramatically increasing from amount of the visitations of web pages. In our previous work, we proposed an algorithm to discover more significant information from visited web pages to provide this information to web designers or policy makers to adjust the presentation of their Web contents. However, this algorithm is time-consuming approach due to it needs to scan the whole database many times. Therefore, we propose a GPGPU-based Preference Utility algorithm to enhance the performance by GPGPU parallel model. The proposed algorithm is developed on NVIDIA CUDA architecture. The experimental results show that the proposed method can achieve about 7x times over CPU-based method. The proposed algorithm can used to mine the information from web log data efficiently.

In this paper, a novel gray-level image-hiding scheme is proposed. The goal of this scheme is to hide multiple important gray-level images into another meaningful gray-level image. The secret images to be protected are first compressed using the vector quantization scheme. Then, the DES cryptosystem is conducted on the VQ indices and related parameters to generate the encrypted message. Finally, the encrypted message is embedded into the rightmost two bits of each pixel in the cover image. According to the experimental results, average image qualities of 44.320 dB and 30.885 dB are achieved for the embedded images and the retrieved secret images, respectively. In other words, multiple secret images can be effectively hidden into one host image of the same size. In addition, the proposed scheme strengthens the protection of the secret images by conducting the DES cryptosystem on the related parameters and the VQ indices of the compressed secret images. Therefore, the proposed scheme provides a secure approach to embed multiple important images into another meaningful image of the same size.

In this paper, a new edge detection scheme based on block truncation coding (BTC) is proposed. As we know, the BTC is a simple and fast scheme for digital image compression. To detect an edge boundary using the BTC scheme, the bit plane information of each BTC-compressed block is exploited, and a simple block type classifier is introduced. The experimental results show that the proposed scheme clearly detects the edge boundaries of digital images while requiring very little computational complexity. Meanwhile, the edge detection process can be incorporated into all BTC variant schemes. In other words, the newly proposed scheme provides a good approach for the detection of edge boundaries using block truncation coding.

The improving technology and the ubiquity of the Internet have allowed more and more people to transmit data via the Internet. The contents of the transmission can be in the form of words, voices, images, or even computer animation. To protect the contents from interceptors' attention, the image hiding technology thus emerged. Some contents transmitted via the Internet can be confidential data such as highly valued product design blueprints or war plans, so it is important to pay more attention to the security of the transmitted data, or what we called secret image in this paper. The point of this paper is to enhance the security of the secret image without causing too much distortion to the host image and in the meantime to shorten the image hiding process time. For better protection, we adopted encryption process DES. In addition, we used greedy algorithm to shorten hiding process and to protect the host image from being severely distorted. To test our proposed method to see whether it indeed achieved its objective, we used two sets of images in our experiment. The results of the experiments showed, when k = 2, our PSNR is close to that of Wang et al.'s optimal LSB substitution, but is not significantly different from that of simple LSB substitution. However, our method took approximately only 1/7 of the time consumed by Wang et al. When k = 3, our PSNR is significantly higher than that of simple LSB substitution. The experimental results confirmed that our method could effectively protect host image quality and shorten the overall hiding time when it enhanced the security of the secret image.

In the past, most image hiding techniques have been applied only to gray scale images. Now, many valuable images are color images. Thus, it has become important to be able to apply image-hiding techniques to hide color images. In this paper, our proposed scheme can not only be applied to "a color host image hiding a color secret image", but also to "a color host image hiding a gray scale secret image". Our scheme utilizes the rightmost 3, 2 and 3 bits of the R, G, B channels of every pixel in the host image to hide related information from the secret image. Meanwhile, we utilize the leftmost 5, 6, 5 bits of the R, G, B channels of every pixel in the host image and set the remaining bits as zero to generate a palette. We then use the palette to conduct color quantization on the secret image to convert its 24-bit pixels into pixels with 8-bit palette index values. DES encryption is then conducted on the index values before the secret image is embedded into the rightmost 3, 2, 3 bits of the R, G, B channels of every pixel in the host image. The experimental results show that even under the worst case scenario our scheme guarantees an average host image PSNR value of 39.184 and an average PSNR value of 27.3415 for the retrieved secret image. In addition to the guarantee of the quality of host images and retrieved secret images, our scheme further strengthens the protection of the secret image by conducting color quantization and DES encryption on the secret image in advance. Therefore, our scheme not only expands the application area of image hiding, but is also practical and secure.