共查询到20条相似文献，搜索用时 375 毫秒
1.
基于发音特征的汉语普通话语音声学建模 总被引：3，自引：0，他引：3
将表征汉语普通话语音特点的发音特征引入汉语普通话语音识别的声学建模中,根据普通话发音特点,确定了用于区别普通话元音、辅音以及声调信息的9种发音特征,并以此为目标值训练神经网络得到语音信号属于各类发音特征的后验概率,将此概率作为语音识别的输入特征建立声学模型。在汉语普通话非特定人大词表自然口语对话识别系统中进行了实验验证,并与基于频谱特征的声学模型进行了比较,在相同解码速度下,由此方法建立的声学模型汉字错误率相对下降6.8%;将发音特征和频谱特征进行了融合实验,融合以后的识别系统相对基于频谱特征系统的汉字错误率相对下降10.1%。上述结果表明,基于发音特征的声学模型更加有效的实现了对语音特性的表征,通过利用发音特征和频谱特征的互补性,能够进一步实现对语音识别性能的提高。 相似文献
2.
摘要为了提高计算机辅助语言学习中自动发音错误检测系统的性能,提出一种声学模型的区分性训练方法。该方法将经过正确度标注的非母语语音数据库上的发音错误检测的F1值的最大化作为模型参数的训练准则。采用Sigmoid 函数对F1值函数进行平滑构造目标函数,并利用构造弱意义辅助函数的方法以及扩展Baum-Welch 形式的参数更新公式进行优化。提出在模型参数更新与音素门限同时优化的策略保证目标函数增长的单调性。发音错误检测实验表明该方法能够有效地增大训练和测试数据检错的F1值。同时训练数据和测试数据上的精确度、召回率以及检测正确度都有明显改进。 相似文献
3.
声波在奥氏体不锈钢焊缝中传播时声束弯曲,为超声成像带来了困难。基于Ogilvy焊缝模型,建立了奥氏体不锈钢焊缝非均匀各向异性声场仿真模型,采用Dijkstra路径搜索算法对各向异性条件下的声传播路径和声传播时间进行了数值模拟和分析。在此基础上,采用Verasonics超声相控阵成像系统,进行了奥氏体不锈钢焊缝的全聚焦成像实验,采集全矩阵回波数据,并结合理论模型计算的声传播路径和相应的传播时间,进行了成像结果修正。结果表明,与均匀介质模型的全聚焦成像结果相比,基于该文模型的焊缝全聚焦成像检测结果具有更高的缺陷定位精度和分辨率,验证了该方法的可行性,为奥氏体不锈钢焊缝成像检测提供了新的思路。 相似文献
4.
5.
分别采用基于数据聚类和基于先验知识的两种研究方法,深入探讨了性别、口音、语速、信道等非语境因素对语音数据分类与建模的影响。为了综合考虑语境、非语境因素在统一框架下建模的问题,采用非语境因素扩展决策树方法。而对于这种方法生成的多套非语境因素相关的高精度声学模型,提出一种依据最大似然准则,动态组合生成测试人相关声学模型的算法。这种方法可以使系统相对误识率平均降低8%~10%。实验结果说明为非语境因素分类建模可以提高声学模型的建模能力,而且模型组合算法可以有效解决统一建模所带来的模型选择问题。 相似文献
6.
针对工业过程控制系统中的故障具有类型多样、时空独立和非线性等特点,使得检测与诊断效率降低,系统性能下降等问题,提出了一种基于自定义多条件约束的多传感器故障检测与诊断机制。该机制,首先考虑了系统的稳态和时空特征建立了非线性过程控制系统多故障模型,并给出了满足条件判定法则;然后对于系统中的单故障,并发故障和通信故障等类型给出了多条件约束法则及独立特性判断;最后提出了通过自定义多条件约束的多传感器故障检测与诊断机制。实验结果表明,在平均检测概率、稳态特征保持能力和系统功耗等方面明显优于无条件约束的机制,可以显著改善过程控制系统性能。 相似文献
7.
作为乳腺癌计算机辅助诊断系统的重要环节,肿块分割的结果严重影响到肿块良恶性的判别.针对现有方法的不足,本文提出了一种基于简化型脉冲耦合神经网络和改进型矢量无边缘活动轮廓模型的乳腺X射线肿块分割方法.首先,通过数学分析计算SPCNN的相关参数与终止条件,进而利用SPCNN模型分割出肿块的初始轮廓.然后,针对传统CV模型的不足,进行相应的修正得到改进型矢量CV模型.最后,结合SPCNN分割出的初始轮廓,利用改进型的矢量CV模型处理ROI分割出肿块.采用北京大学人民医院乳腺中心提供的临床图像以及DDSM数据库的图像进行对比实验,实验结果表明,本文方法相比较现有方法分割结果更为准确,尤其是在处理东方女性致密性案例时,本文方法更有优势. 相似文献
8.
物理实验中应用计算机辅助教学的分析与思考 总被引：1，自引：0，他引：1
本文简要分析了物理实验教学的特点,依据实践经验探讨了计算机辅助教学在物理产验中具体的应用方向,研究了应用计算机辅助教学可能出现的问题。 相似文献
9.
种子纯度反映种子品种在特征特性方面典型一致的程度,提高种子纯度检测的准确性和可靠性对保证种子的质量具有重要的意义。高光谱图像技术可以同时反映种子的内部特征和外部特征,在农产品无损检测中已经得到广泛应用。利用近红外高光谱图像实现农产品无损检测的实质就是建立光谱信息与农产品品质参数之间的数学模型关系。但光谱信息易受环境、时间的影响,当待测样本的产地或者年份发生改变时光谱信息也随之改变,导致建立的模型的稳定性变差、泛化能力减弱。针对这一问题,采用主动学习算法选择具有代表性的待测样本,最终以添加最少最优的样本数来扩大原模型的样本空间,从而实现模型的快速更新,提高模型的稳定性,同时与基于随机选择算法（RS）和Kennard-Stone算法(KS)的模型更新效果进行比较。实验结果表明：在不同样本集划分比例下（1∶1, 3∶1, 4∶1）,利用主动学习添加40个新样本更新后的2010年的玉米种子纯度检测模型对2011年新样本的预测精度由47%,33.75%,49%提高到98.89%,98.33%,98.33%;利用主动学习添加56个新样本更新后的2011年的玉米种子纯度检测模型对2010年新样本的预测精度由50.83%,54.58%,53.75%提高到94.57%,94.02%,94.57%;同时基于主动学习算法的模型更新效果明显优于RS和KS。因此基于主动学习算法实现玉米种子纯度检测模型的更新是可行的。 相似文献
10.
近红外光谱分析技术依赖于表征光谱向量和预测目标之间关系的化学计量学方法。然而,样品的光谱由信号和各种噪声组成,传统化学计量学方法较难直接提取光谱的有效特征,并为复杂的预测任务建立具有较强泛用性的校正模型。进一步地,受限于仪器间的差异,在一台仪器上建立的模型应用于另一台仪器时,难以取得相同的定量分析结果。为此,提出了一种基于卷积神经网络和迁移学习的定量分析建模及模型传递方案,以提高模型在单仪器和跨仪器上的预测性能。在卷积神经网络的基础上,一种结合多尺度特征融合和残差结构,名为MSRCNN的先进模型被设计,并在主仪器上展现了卓越的预测能力。然后,设计了四种的基于fine-tune模型迁移策略,将在主仪器上建立的MSRCNN模型迁移到从仪器。在药品和小麦的公开数据集上的实验结果表明,MSRCNN在主仪器上的RMSE和R2分别为2.587,0.981和0.309,0.977,优于PLS,SVM和CNN。在利用30个从仪器的样本微调主仪器建立的模型后,迁移MSRCNN中的卷积层和全连接层的方案取得了最好效果,其RMSE和R2可分别达到2.289,0.982和0.379,0.965。增加参与模型微调的从仪器样本,可进一步提高性能。 相似文献
11.
常见的多说话人语音合成有参数自适应及添加说话人标签两种方法。参数自适应方法获得的模型仅支持合成经过自适应的说话人的语音,模型不够鲁棒。传统的添加说话人标签的方法需要有监督地获得语音的说话人信息,并没有从语音信号本身无监督地学习说话人标签。为解决这些问题,提出了一种基于变分自编码器的自回归多说话人语音合成方法。方法首先利用变分自编码器无监督地学习说话人的信息并将其隐式编码为说话人标签,之后与文本的语言学特征送入到一个自回归声学参数预测网络中。此外,为了抑制多说话人语音数据引起的基频预测过拟合问题,声学参数网络采用了基频多任务学习的方法。预实验表明,自回归结构的加入降低了频谱误差1.018 dB,基频多任务学习降低了基频均方根误差6.861 Hz。在后续的多说话人对比实验中,提出的方法在3个多说话人实验的平均主观意见分(MOS)打分上分别达到3.71,3.55,3.15,拼音错误率分别为6.71%,7.54%,9.87%,提升了多说话人语音合成的音质。 相似文献
12.
Julie Wang Alexander Wood Chao Gao Kayvan Najarian Jonathan Gryak 《Entropy (Basel, Switzerland)》2021,23(4)
The spleen is one of the most frequently injured organs in blunt abdominal trauma. Computed tomography (CT) is the imaging modality of choice to assess patients with blunt spleen trauma, which may include lacerations, subcapsular or parenchymal hematomas, active hemorrhage, and vascular injuries. While computer-assisted diagnosis systems exist for other conditions assessed using CT scans, the current method to detect spleen injuries involves the manual review of scans by radiologists, which is a time-consuming and repetitive process. In this study, we propose an automated spleen injury detection method using machine learning. CT scans from patients experiencing traumatic injuries were collected from Michigan Medicine and the Crash Injury Research Engineering Network (CIREN) dataset. Ninety-nine scans of healthy and lacerated spleens were split into disjoint training and test sets, with random forest (RF), naive Bayes, SVM, k-nearest neighbors (k-NN) ensemble, and subspace discriminant ensemble models trained via 5-fold cross validation. Of these models, random forest performed the best, achieving an Area Under the receiver operating characteristic Curve (AUC) of 0.91 and an F1 score of 0.80 on the test set. These results suggest that an automated, quantitative assessment of traumatic spleen injury has the potential to enable faster triage and improve patient outcomes. 相似文献
13.
HOU Chunjia;HE Boxia;HU Jinsong;YU Jie;CHEN Xuyang 《光子学报》2024,53(3):144-152
The characteristic size and surface quality of O-ring seals （hereinafter referred to as “O-ring”） used in aerospace and guided weapon systems are important factors affecting the reliability of the main engine, and must be 100% fully checked. Due to the flexible characteristics of the O-ring material and the omnidirectional curved surface feature of the outer surface, the current manual measurement and detection methods have three major drawbacks： low efficiency, unstable results and high manpower consumption, which can no longer meet the requirements of the rapid development of aerospace and defense industries. With the advent of convolutional neural networks, target detection algorithms based on deep learning are widely used in the field of target detection because of their simple structure and good versatility. The micro-aerospace the O-ring studied in this paper has an inner diameter size range of Φ 1.8 mm-Φ 20 mm. Through the analysis of the surface topography of the defects, it is found that most of the defects have the characteristics of tiny targets and the pixels of the marked defects are less than 0.33% of the total pixels of the image, which is a typical tiny target detection. Compared with other computer vision tasks, tiny target detection has problems like fewer available features, higher positioning accuracy requirements, lower proportions of tiny targets in datasets, sample imbalance and tiny target aggregation. Because of its omnidirectional curved surface features, the O-ring presents severe bright areas and dark areas in images from any angle. The random defects are intertwined with these non-uniform areas which causes great difficulties in the detection and classification of surface defects. Especially for micro-O-ring, tiny defects impose higher demands on algorithm sensitivity and classification ability. Although the target detection algorithm based on deep learning has good detection capability but its detection efficiency is relatively low, and there is room for improvement in detection accuracy.To address the above problems, two deep-learning-based algorithms are proposed for detecting surface defects on the O-ring. By adding multi-head attention mechanisms to the inverse residual blocks of MobileNetv2, we constructed a lightweight backbone network called Efficient Model. By using the Next Hybrid strategy, we fused multiple attention mechanism modules from the industrial-grade Transformer network to build a Next Generation Vision Transformer backbone network. In each of these two backbone networks, feature extraction networks were added to design the Efficient-FPN Model and Transformer-FPN Model detection algorithms. The experimental results show that the mAP of the Efficient-FPN Model and Transformer-FPN Model detection algorithms is higher than that of YOLOv5s, YOLOv5x and YOLOv5z, among which the mAP of the Transformer-FPN model is the highest, reaching 91.4%. The Efficient-FPN Model has the fastest detection speed of the five models, reaching 110.8 frame/s. The mAP of the Efficient-FPN Model reached 86.1%, which was also higher than other YOLOv5 algorithms, and it was the detection model with the best comprehensive performance. The above algorithm is deployed in the self-developed intelligent measurement and inspection equipment of aerospace seal ring, and the purpose of detecting omnidirectional curved flexible parts is realized quickly and accurately. 相似文献
14.
Shadow is one of the fundamental indicators of remote sensing image which could cause loss or interference of the target data. As a result, the detection and removal of shadow has already been the hotspot of current study because of the complicated background information. In the following passage, a model combining the Atmospheric Transport Model (hereinafter abbreviated as ATM) with the Poisson Equation, AP ShadowNet, is proposed for the shadow detection and removal of remote sensing images by unsupervised learning. This network based on a preprocessing network based on ATM, A Net, and a network based on the Poisson Equation, P Net. Firstly, corresponding mapping between shadow and unshaded area is generated by the ATM. The brightened image will then enter the Confrontation identification in the P Net. Lastly, the reconstructed image is optimized on color consistency and edge transition by Poisson Equation. At present, most shadow removal models based on neural networks are significantly data-driven. Fortunately, by the model in this passage, the unsupervised shadow detection and removal could be released from the data source restrictions from the remote sensing images themselves. By verifying the shadow removal on our model, the result shows a satisfying effect from a both qualitative and quantitative angle. From a qualitative point of view, our results have a prominent effect on tone consistency and removal of detailed shadows. From the quantitative point of view, we adopt the non-reference evaluation indicators: gradient structure similarity (NRSS) and Natural Image Quality Evaluator (NIQE). Combining various evaluation factors such as reasoning speed and memory occupation, it shows that it is outstanding among other current algorithms. 相似文献
15.
16.
Future communication networks must address the scarce spectrum to accommodate extensive growth of heterogeneous wireless devices. Efforts are underway to address spectrum coexistence, enhance spectrum awareness, and bolster authentication schemes. Wireless signal recognition is becoming increasingly more significant for spectrum monitoring, spectrum management, secure communications, among others. Consequently, comprehensive spectrum awareness on the edge has the potential to serve as a key enabler for the emerging beyond 5G (fifth generation) networks. State-of-the-art studies in this domain have (i) only focused on a single task – modulation or signal (protocol) classification – which in many cases is insufficient information for a system to act on, (ii) consider either radar or communication waveforms (homogeneous waveform category), and (iii) does not address edge deployment during neural network design phase. In this work, for the first time in the wireless communication domain, we exploit the potential of deep neural networks based multi-task learning (MTL) framework to simultaneously learn modulation and signal classification tasks while considering heterogeneous wireless signals such as radar and communication waveforms in the electromagnetic spectrum. The proposed MTL architecture benefits from the mutual relation between the two tasks in improving the classification accuracy as well as the learning efficiency with a lightweight neural network model. We additionally include experimental evaluations of the model with over-the-air collected samples and demonstrate first-hand insight on model compression along with deep learning pipeline for deployment on resource-constrained edge devices. We demonstrate significant computational, memory, and accuracy improvement of the proposed model over two reference architectures. In addition to modeling a lightweight MTL model suitable for resource-constrained embedded radio platforms, we provide a comprehensive heterogeneous wireless signals dataset for public use. 相似文献
17.
18.
Insider threats are malicious acts that can be carried out by an authorized employee within an organization. Insider threats represent a major cybersecurity challenge for private and public organizations, as an insider attack can cause extensive damage to organization assets much more than external attacks. Most existing approaches in the field of insider threat focused on detecting general insider attack scenarios. However, insider attacks can be carried out in different ways, and the most dangerous one is a data leakage attack that can be executed by a malicious insider before his/her leaving an organization. This paper proposes a machine learning-based model for detecting such serious insider threat incidents. The proposed model addresses the possible bias of detection results that can occur due to an inappropriate encoding process by employing the feature scaling and one-hot encoding techniques. Furthermore, the imbalance issue of the utilized dataset is also addressed utilizing the synthetic minority oversampling technique (SMOTE). Well known machine learning algorithms are employed to detect the most accurate classifier that can detect data leakage events executed by malicious insiders during the sensitive period before they leave an organization. We provide a proof of concept for our model by applying it on CMU-CERT Insider Threat Dataset and comparing its performance with the ground truth. The experimental results show that our model detects insider data leakage events with an AUC-ROC value of 0.99, outperforming the existing approaches that are validated on the same dataset. The proposed model provides effective methods to address possible bias and class imbalance issues for the aim of devising an effective insider data leakage detection system. 相似文献
19.
With the quick development of sensor technology in recent years, online detection of early fault without system halt has received much attention in the field of bearing prognostics and health management. While lacking representative samples of the online data, one can try to adapt the previously-learned detection rule to the online detection task instead of training a new rule merely using online data. As one may come across a change of the data distribution between offline and online working conditions, it is challenging to utilize the data from different working conditions to improve detection accuracy and robustness. To solve this problem, a new online detection method of bearing early fault is proposed in this paper based on deep transfer learning. The proposed method contains an offline stage and an online stage. In the offline stage, a new state assessment method is proposed to determine the period of the normal state and the degradation state for whole-life degradation sequences. Moreover, a new deep dual temporal domain adaptation (DTDA) model is proposed. By adopting a dual adaptation strategy on the time convolutional network and domain adversarial neural network, the DTDA model can effectively extract domain-invariant temporal feature representation. In the online stage, each sequentially-arrived data batch is directly fed into the trained DTDA model to recognize whether an early fault occurs. Furthermore, a health indicator of target bearing is also built based on the DTDA features to intuitively evaluate the detection results. Experiments are conducted on the IEEE Prognostics and Health Management (PHM) Challenge 2012 bearing dataset. The results show that, compared with nine state-of-the-art fault detection and diagnosis methods, the proposed method can get an earlier detection location and lower false alarm rate. 相似文献