针对6G时代将会是移动通信与人工智能紧密结合的时代,产生数量庞大的边缘智能信号处理节点的趋势,提出了一种可部署于资源受限的边缘设备上的高效智能电磁信号识别模型。首先,通过绘制电磁信号的星座图将电磁信号具象为二维图像,并根据...针对6G时代将会是移动通信与人工智能紧密结合的时代,产生数量庞大的边缘智能信号处理节点的趋势,提出了一种可部署于资源受限的边缘设备上的高效智能电磁信号识别模型。首先,通过绘制电磁信号的星座图将电磁信号具象为二维图像,并根据归一化点密度对星座图上色以实现特征增强;然后,使用二值化深度神经网络对其进行识别,在保证识别准确率的同时明显降低了模型存储开销以及计算开销。采用电磁信号调制识别问题进行验证,实验选取常用的8种数字调制信号,选择加性高斯白噪声为信道环境。实验结果表明,所提方案可以在信噪比为-6~6 d B的噪声条件下获得96.1%的综合识别率,网络模型大小仅为166 KB,部署于树莓派4B的执行时间为290 ms,相比于同规模的全精度网络,准确率提升了0.6%,模型缩减到1/26.16,运行时间缩减到1/2.37。展开更多
针对目前BNN(Binarized Neural Network)剪枝方法存在剪枝比例低、识别准确率显著下降以及依赖训练后微调的问题,提出了一种基于三值向二值演化的滤波器级的BNN剪枝方法,命名为ETB(Evolution from Ternary to Binary)。ETB是基于学习的...针对目前BNN(Binarized Neural Network)剪枝方法存在剪枝比例低、识别准确率显著下降以及依赖训练后微调的问题,提出了一种基于三值向二值演化的滤波器级的BNN剪枝方法,命名为ETB(Evolution from Ternary to Binary)。ETB是基于学习的,通过在BNN的量化函数中引入可训练的量化阈值,使权重和激活值逐渐从三值演化到二值或零,旨在使网络在训练期间自动识别不重要的结构。此外,一个剪枝率调节算法也被设计用于调控网络的剪枝率。训练后,全零滤波器和对应的输出通道可被直接裁剪而获得精简的BNN,无需微调。为证明提出方法的可行性和其提升BNN推理效率而不牺牲准确率的潜力,在CIFAR-10上进行实验:在CIFAR-10数据集上,ETB对VGG-Small模型进行了46.3%的剪枝,模型大小压缩至0.34 MByte,准确率为89.97%,并在ResNet-18模型上进行了30.01%的剪枝,模型大小压缩至1.33 MByte,准确率为90.79%。在准确率和参数量方面,对比一些现有的BNN剪枝方法,ETB具有一定的优势。展开更多
This paper proposes a new method for ship recognition and classification using sound produced and radiated underwater. To do so, a three-step procedure is proposed. First, the preprocessing operations are utilized to ...This paper proposes a new method for ship recognition and classification using sound produced and radiated underwater. To do so, a three-step procedure is proposed. First, the preprocessing operations are utilized to reduce noise effects and provide signal for feature extraction. Second, a binary image, made from frequency spectrum of signal segmentation, is formed to extract effective features. Third, a neural classifier is designed to classify the signals. Two approaches, the proposed method and the fractal-based method are compared and tested on real data. The comparative results indicated better recognition ability and more robust performance of the proposed method than the fractal-based method. Therefore, the proposed method could improve the recognition accuracy of underwater acoustic targets.展开更多
文摘针对6G时代将会是移动通信与人工智能紧密结合的时代,产生数量庞大的边缘智能信号处理节点的趋势,提出了一种可部署于资源受限的边缘设备上的高效智能电磁信号识别模型。首先,通过绘制电磁信号的星座图将电磁信号具象为二维图像,并根据归一化点密度对星座图上色以实现特征增强;然后,使用二值化深度神经网络对其进行识别,在保证识别准确率的同时明显降低了模型存储开销以及计算开销。采用电磁信号调制识别问题进行验证,实验选取常用的8种数字调制信号,选择加性高斯白噪声为信道环境。实验结果表明,所提方案可以在信噪比为-6~6 d B的噪声条件下获得96.1%的综合识别率,网络模型大小仅为166 KB,部署于树莓派4B的执行时间为290 ms,相比于同规模的全精度网络,准确率提升了0.6%,模型缩减到1/26.16,运行时间缩减到1/2.37。
文摘针对目前BNN(Binarized Neural Network)剪枝方法存在剪枝比例低、识别准确率显著下降以及依赖训练后微调的问题,提出了一种基于三值向二值演化的滤波器级的BNN剪枝方法,命名为ETB(Evolution from Ternary to Binary)。ETB是基于学习的,通过在BNN的量化函数中引入可训练的量化阈值,使权重和激活值逐渐从三值演化到二值或零,旨在使网络在训练期间自动识别不重要的结构。此外,一个剪枝率调节算法也被设计用于调控网络的剪枝率。训练后,全零滤波器和对应的输出通道可被直接裁剪而获得精简的BNN,无需微调。为证明提出方法的可行性和其提升BNN推理效率而不牺牲准确率的潜力,在CIFAR-10上进行实验:在CIFAR-10数据集上,ETB对VGG-Small模型进行了46.3%的剪枝,模型大小压缩至0.34 MByte,准确率为89.97%,并在ResNet-18模型上进行了30.01%的剪枝,模型大小压缩至1.33 MByte,准确率为90.79%。在准确率和参数量方面,对比一些现有的BNN剪枝方法,ETB具有一定的优势。
文摘This paper proposes a new method for ship recognition and classification using sound produced and radiated underwater. To do so, a three-step procedure is proposed. First, the preprocessing operations are utilized to reduce noise effects and provide signal for feature extraction. Second, a binary image, made from frequency spectrum of signal segmentation, is formed to extract effective features. Third, a neural classifier is designed to classify the signals. Two approaches, the proposed method and the fractal-based method are compared and tested on real data. The comparative results indicated better recognition ability and more robust performance of the proposed method than the fractal-based method. Therefore, the proposed method could improve the recognition accuracy of underwater acoustic targets.
