The broadband frequency output of gravitational-wave (GW) detectors is a non-stationary and non-Gaussian time series data stream dominated by noise populated by local disturbances and transient artifacts, which evolve...
详细信息
The broadband frequency output of gravitational-wave (GW) detectors is a non-stationary and non-Gaussian time series data stream dominated by noise populated by local disturbances and transient artifacts, which evolve on the same timescale as the GW signals and may corrupt the astrophysical information. We study a denoising algorithm dedicated to expose the astrophysical signals by employing a convolutional neural network in the encoder-decoder configuration, i.e. apply the denoising procedure of coalescing binary black hole signals to the publicly available LIGO O1 time series strain data. The denoising convolutional autoencoder neural network is trained on a dataset of simulated astrophysical signals injected into the real detector's noise and a dataset of detector noise artifacts ('glitches'), and its fidelity is tested on real GW events from O1 and O2 LIGO-Virgo observing runs.
The parametric Bayesian Feature Enhancement (BFE) and a datadriven denoising autoencoder (DA) both bring performance gains in severe single-channel speech recognition conditions. The first can be adjusted to different...
详细信息
ISBN:
(纸本)9781467369985
The parametric Bayesian Feature Enhancement (BFE) and a datadriven denoising autoencoder (DA) both bring performance gains in severe single-channel speech recognition conditions. The first can be adjusted to different conditions by an appropriate parameter setting, while the latter needs to be trained on conditions similar to the ones expected at decoding time, making it vulnerable to a mismatch between training and test conditions. We use a DNN backend and study reverberant ASR under three types of mismatch conditions: different room reverberation times, different speaker to microphone distances and the difference between artificially reverberated data and the recordings in a reverberant environment. We show that for these mismatch conditions BFE can provide the targets for a DA. This unsupervised adaptation provides a performance gain over the direct use of BFE and even enables to compensate for the mismatch of real and simulated reverberant data.
Brain tumours are one of the common diseases in human beings. Currently, brain Nuclear Magnetic Resonance(MRI) is the main means of detecting brain diseases. There are many works related to the classification and nois...
详细信息
Brain tumours are one of the common diseases in human beings. Currently, brain Nuclear Magnetic Resonance(MRI) is the main means of detecting brain diseases. There are many works related to the classification and noise reduction of brain MRI images, but there are few works that combine these two parts simultaneously, i.e., noise reduction processing and classification of brain images at the same time. In this work, these two parts of work will be combined together, extracting the features of brain MRI images through an encoder composed of convolutional neural networks, using the features to classify the images, and then using the features to reduce the noise of the images to produce a low-noise image. According to the model,this work also proposes a new loss function. The new loss function is composed by adjusting the weights of the loss functions related to the classification and denoising tasks. This paper tries three different ways of weight assignment, and the experimental results show that the dynamic parameter assignment approach achieves the best results for image classification, but none of the three approaches achieves acceptable results for noise reduction.
This paper investigates a multi-channel denoising autoencoder (DAE)-based speech enhancement approach. In recent years, deep neural network (DNN)-based monaural speech enhancement and robust automatic speech recogniti...
详细信息
In this paper, we propose a robust distant-talking speech recognition by combining cepstral domain denoising autoencoder (DAE) and temporal structure normalization (TSN) filter. For the proposed method, after applying...
详细信息
ISBN:
(纸本)9781479942190
In this paper, we propose a robust distant-talking speech recognition by combining cepstral domain denoising autoencoder (DAE) and temporal structure normalization (TSN) filter. For the proposed method, after applying a DAE in the cepstral domain of speech to suppress reverberation, we apply a postprocessing technology based on temporal structure normalization (TSN) filter to reduce the noise and reverberation effects by normalizing the modulation spectra to reference spectra of clean speech. The proposed method was evaluated using speech in simulated and real reverberant environments. By combining a cepstral-domain DAE and TSN, the average Word Error Rate (WER) was reduced from 25.2% of the baseline system to 21.2% in simulated environments and from 47.5% to 41.3% in real environments, respectively.
denoising autoencoder (DAE) is effective in restoring clean speech from noisy observations. In addition, it is easy to be stacked to a deep denoising autoencoder (DDAE) architecture to further improve the performance....
详细信息
ISBN:
(纸本)9781634394352
denoising autoencoder (DAE) is effective in restoring clean speech from noisy observations. In addition, it is easy to be stacked to a deep denoising autoencoder (DDAE) architecture to further improve the performance. In most studies, it is supposed that the DAE or DDAE can learn any complex transform functions to approximate the transform relation between noisy and clean speech. However, for large variations of speech patterns and noisy environments, the learned model is lack of focus on local transformations. In this study, we propose an ensemble modeling of DAE to learn both the global and local transform functions. In the ensemble modeling, local transform functions are learned by several DAEs using data sets obtained from unsupervised data clustering and partition. The final transform function used for speech restoration is a combination of all the learned local transform functions. Speech denoising experiments were carried out to examine the performance of the proposed method. Experimental results showed that the proposed ensemble DAE model provided superior restoration accuracy than traditional DAE models.
The denoising autoencoder (DAE) has been successfully applied to acoustic emotion recognition lately. In this paper, we adopt the framework of the modified DAE introduced in [1] that projects the input signal to two d...
详细信息
ISBN:
(纸本)9781479928934
The denoising autoencoder (DAE) has been successfully applied to acoustic emotion recognition lately. In this paper, we adopt the framework of the modified DAE introduced in [1] that projects the input signal to two different hidden representations, for neutral and emotional speech respectively, and uses the emotional representation for the classification task. We propose to model gender information for more robust emotional representation in this work. For neutral representation, male and female dependent DAEs are built using non-emotional speech with the aim of capturing distinct information between the two genders. The emotional hidden representation is shared for the two genders in order to model more emotion specific characteristics, and is used as features in a back-end classifier for emotion recognition. We propose different optimization objectives in training the DAEs. Our experimental results show improvement on unweighted accuracy compared with previous work using the modified DAE method and the classifiers using the standard static features. Further performance gain can be achieved by structural level system combination.
Surface electromyography (sEMG) is a widely employed bio-signal that captures human muscle activity via electrodes placed on the skin. Several studies have proposed methods to remove sEMG contaminants, as non-invasive...
详细信息
Surface electromyography (sEMG) is a widely employed bio-signal that captures human muscle activity via electrodes placed on the skin. Several studies have proposed methods to remove sEMG contaminants, as non-invasive measurements render sEMG susceptible to various contaminants. However, these approaches often rely on heuristic-based optimization and are sensitive to the contaminant type. A more potent, robust, and generalized sEMG denoising approach should be developed for various healthcare and human-computer interaction applications. This paper proposes a novel neural network (NN)-based sEMG denoising method called TrustEMG-Net. It leverages the potent nonlinear mapping capability and data-driven nature of NNs. TrustEMG-Net adopts a denoising autoencoder structure by combining U-Net with a Transformer encoder using a representation-masking approach. The proposed approach is evaluated using the Ninapro sEMG database with five common contamination types and signal-to-noise ratio (SNR) conditions. Compared with existing sEMG denoising methods, TrustEMG-Net achieves exceptional performance across the five evaluation metrics, exhibiting a minimum improvement of 20%. Its superiority is consistent under various conditions, including SNRs ranging from -14 to 2 dB and five contaminant types. An ablation study further proves that the design of TrustEMG-Net contributes to its optimality, providing high-quality sEMG and serving as an effective, robust, and generalized denoising solution for sEMG applications.
The emergence of fifth-generation (5G) mobile communication technologies has propelled the advancement of the internet of things (IoT). Nevertheless, the intricate nature of the IoT mobile communication environment an...
详细信息
The emergence of fifth-generation (5G) mobile communication technologies has propelled the advancement of the internet of things (IoT). Nevertheless, the intricate nature of the IoT mobile communication environment and the fluctuating characteristics of the signal's present substantial obstacles to current spectrum detection techniques for future communication. Hence, an artificial intelligent spectrum sensing technique is introduced, which integrates artificial intelligent, IoT and denoising autoencoder (DAE) with an enhanced long-short-term memory (LSTM) neural network. The DAE utilises encoding and decoding to retrieve the fundamental structural characteristics of mobile signals, while the enhanced LSTM spectrum sensing classifier model incorporates previous moment information features to classify the time-series signal sequences. This method has demonstrated a 45% improvement in perception performance compared to SVM, RNN, LeNet5, LVQ, and Elman algorithm.
Stealthy false data injection attack (FDIA) that intentionally modifies measurement data of smart grid meters to bypass the traditional bad data detection module is one of menacing cyber attacks in smart grid. Due to ...
详细信息
Stealthy false data injection attack (FDIA) that intentionally modifies measurement data of smart grid meters to bypass the traditional bad data detection module is one of menacing cyber attacks in smart grid. Due to requiring no costly labeling abnormal measurement data, deep neural networks (DNNs)-based unsupervised FDIA detection has attracted great attentions. However, the existing schemes have two weaknesses. First, most schemes did not take into account the inherent spatial relationships between measurements in the grid. Second, for practical usage, the robustness and generalization of the trained FDIA detection scheme will be influenced by potential noisy measurement data. To address the issues above, based on the spatial graph neural network (GNN) architecture, a novel FDIA detection and localization scheme is proposed, named as recursive variational graph autoencoder (ReVGAE). Specifically, our contributions are following. The variational graph autoencoder (VGAE) module in our proposed ReVGAE innovatively plays dual roles: 1) data and topology reconstructor and 2) the denoising module. The first role aims to simultaneously reconstruct both nodes' temporal measurements and topological relationship between nodes. In the second role, the outputs of VGAE as the reconstructor (i.e., the reconstructed temporal measurements) are intentionally used as the artificially noisy samples, and recursively fed into VGAE as input to improve the model's robustness. Then, the residual between the finally reconstructed and the observed measurement data on each node is viewed as anomaly score to judge whether FDIA temporally happens on each node. Thorough experiments on a real grid system demonstrate that the proposed ReVGAE outperforms other variational autoencoder and GNN-based FDIA anomaly detection schemes.
暂无评论