模拟信源标量量化的模糊c-均值算法

为改进模拟信源标量量化的收敛速度,将模糊逻辑中的模糊c-均值算法进行适当改造,应用到模拟信源的标量量化过程中,即形成了模拟信源标量量化的模糊c-均值算法。算法将迭代过程中的分区矩阵“元素化”放宽了限制条件。仿真结果表明,新的模糊c-均值算法在选择合适的精度E的条件下,随着分区数目增加和训练序列长度的增加,收敛速度要远优于传统的c-均值算法,同时通过控制参数E,可以在收敛速度和量化失真度量M SE之间进行权衡。

Fuzzy c-mean algorithm on scalar quantization of analog source

To improve the convergence rate of the scalar quantization for analog source,the fuzzy c-mean algorithm(FCM) in fuzzy logic was applied to the scalar quantization of analog source through some alteration.The algorithm relaxed the restricted conditions by fuzzing the element of partition matrix in the iterative process.The simulations indicated that the convergence rate of FCM was far more excellent than conventional c-mean algorithm in increasing the number of partition and the length of trained sequence with appropriate E.And it could be a tradeoff between the convergence rate and the quantization distortion measure MSE(mean-square error) by controlling E.