At present,the traditional channel estimation algorithms have the disadvantages of over-reliance on initial conditions and high complexity.The bacterial foraging optimization(BFO)-based algorithm has been applied in w...At present,the traditional channel estimation algorithms have the disadvantages of over-reliance on initial conditions and high complexity.The bacterial foraging optimization(BFO)-based algorithm has been applied in wireless communication and signal processing because of its simple operation and strong self-organization ability.But the BFO-based algorithm is easy to fall into local optimum.Therefore,this paper proposes the quantum bacterial foraging optimization(QBFO)-binary orthogonal matching pursuit(BOMP)channel estimation algorithm to the problem of local optimization.Firstly,the binary matrix is constructed according to whether atoms are selected or not.And the support set of the sparse signal is recovered according to the BOMP-based algorithm.Then,the QBFO-based algorithm is used to obtain the estimated channel matrix.The optimization function of the least squares method is taken as the fitness function.Based on the communication between the quantum bacteria and the fitness function value,chemotaxis,reproduction and dispersion operations are carried out to update the bacteria position.Simulation results showthat compared with other algorithms,the estimationmechanism based onQBFOBOMP algorithm can effectively improve the channel estimation performance of millimeter wave(mmWave)massive multiple input multiple output(MIMO)systems.Meanwhile,the analysis of the time ratio shows that the quantization of the bacteria does not significantly increase the complexity.展开更多
针对各种环境声音对声音事件识别的影响,提出一种基于优化的正交匹配追踪(OOMP)和深度置信网(DBN)的声音事件识别方法。首先,利用粒子群优化(PSO)算法优化OMP稀疏分解,在实现正交匹配追踪(OMP)的快速稀疏分解的同时,保留声音信号的主体...针对各种环境声音对声音事件识别的影响,提出一种基于优化的正交匹配追踪(OOMP)和深度置信网(DBN)的声音事件识别方法。首先,利用粒子群优化(PSO)算法优化OMP稀疏分解,在实现正交匹配追踪(OMP)的快速稀疏分解的同时,保留声音信号的主体部分,抑制噪声对声音信号的影响;接着,对重构声音信号提取Mel频率倒谱系数(MFCC)、OMP时-频特征和基音频率(Pitch)特征,组成OOMP的复合特征;最后,使用DBN对提取的OOMP特征进行特征学习,并对40种声音事件在不同环境不同信噪比下进行识别。实验结果表明,OOMP特征结合DBN的方法适用于各种环境声下的声音事件识别,而且能有效地识别各种环境下的声音事件,即使在信噪比(SNR)为0 d B的情况下,仍然能保持平均60%的识别率。展开更多
基金supported by the National Natural Science Foundation of China(Nos.61861015,62061013 and 61961013)Key Research and Development Program of Hainan Province(No.ZDYF2019011)+3 种基金National Key Research and Development Program of China(No.2019CXTD400)Young Elite Scientists Sponsorship Program by CAST(No.2018QNRC001)Scientific Research Setup Fund of Hainan University(No.KYQD(ZR)1731)the Natural Science Foundation High-Level Talent Project of Hainan Province(No.622RC619).
文摘At present,the traditional channel estimation algorithms have the disadvantages of over-reliance on initial conditions and high complexity.The bacterial foraging optimization(BFO)-based algorithm has been applied in wireless communication and signal processing because of its simple operation and strong self-organization ability.But the BFO-based algorithm is easy to fall into local optimum.Therefore,this paper proposes the quantum bacterial foraging optimization(QBFO)-binary orthogonal matching pursuit(BOMP)channel estimation algorithm to the problem of local optimization.Firstly,the binary matrix is constructed according to whether atoms are selected or not.And the support set of the sparse signal is recovered according to the BOMP-based algorithm.Then,the QBFO-based algorithm is used to obtain the estimated channel matrix.The optimization function of the least squares method is taken as the fitness function.Based on the communication between the quantum bacteria and the fitness function value,chemotaxis,reproduction and dispersion operations are carried out to update the bacteria position.Simulation results showthat compared with other algorithms,the estimationmechanism based onQBFOBOMP algorithm can effectively improve the channel estimation performance of millimeter wave(mmWave)massive multiple input multiple output(MIMO)systems.Meanwhile,the analysis of the time ratio shows that the quantization of the bacteria does not significantly increase the complexity.
文摘针对各种环境声音对声音事件识别的影响,提出一种基于优化的正交匹配追踪(OOMP)和深度置信网(DBN)的声音事件识别方法。首先,利用粒子群优化(PSO)算法优化OMP稀疏分解,在实现正交匹配追踪(OMP)的快速稀疏分解的同时,保留声音信号的主体部分,抑制噪声对声音信号的影响;接着,对重构声音信号提取Mel频率倒谱系数(MFCC)、OMP时-频特征和基音频率(Pitch)特征,组成OOMP的复合特征;最后,使用DBN对提取的OOMP特征进行特征学习,并对40种声音事件在不同环境不同信噪比下进行识别。实验结果表明,OOMP特征结合DBN的方法适用于各种环境声下的声音事件识别,而且能有效地识别各种环境下的声音事件,即使在信噪比(SNR)为0 d B的情况下,仍然能保持平均60%的识别率。
