排序方式: 共有4条查询结果,搜索用时 0 毫秒
1
1.
在低信噪比情况下,该文提出一种新的针对正交频分复用(OFDM)系统信道阶数和噪声方差的非数据辅助(NDA)估计算法。算法中应用了一种新的基于联合极大几何均值(MGM)的代价函数。新的代价函数不仅利用了循环前缀(CP)冗余性,同时也利用了信道记忆性。对比只利用了CP的方法,该算法可以在低信噪比情况下更准确地估计信道阶数和噪声方差。仿真结果表明,在低信噪比情况下,该算法针对信道阶数的估计得到约10 dB的信噪比增益;同时,对噪声方差的估计,该算法显著提高了估计精度,抑制了信噪比20 dB以下估计性能恶化的现象。 相似文献
2.
3.
针对超声乳腺图像中对不同规模病变的分割鲁棒性不足的问题,提出了一种多尺度注意力小波网络(MAW-Net).通过设计两个轻量的网络模块,多尺度拼接模块和跳过连接升维模块,达到在不同尺度上集成丰富的特征和全局上下文信息,减少编码器和解码器之间的语义差距以适应不同规模病变分割的目的.并引入双树复小波变换,很好地削弱了噪声影响.在两个公共的乳腺超声数据集UDIAT和BUSI数据集上进行测试,其Dice系数分别达到91.32%和84.23%.并与其他6种先进的图像分割方法进行比较,具备强分割鲁棒性、噪声影响小等优势. 相似文献
4.
Fountain codes provide an efficient way to transfer information over erasure channels like the Internet. LT codes are the first codes fully realizing the digital fountain concept. They are asymptotically optimal rateless erasure codes with highly efficient encoding and decoding algorithms. In theory, for each encoding symbol of LT codes, its degree is randomly chosen according to a predetermined degree distribution, and its neighbours used to generate that encoding symbol are chosen uniformly at random. Practical implementation of LT codes usually realizes the randomness through pseudo-randomness number generator like linear congruential method. This paper applies the pseudo-randomness of chaotic sequence in the implementation of LT codes. Two Kent chaotic maps are used to determine the degree and neighbour(s) of each encoding symbol. It is shown that the implemented LT codes based on chaos perform better than the LT codes implemented by the traditional pseudo-randomness number generator. 相似文献
1