论文部分内容阅读
本文提出一种能进行数据变换、量化和编码的神经网络模型,引入离散阶梯状量化函数作为隐层神经元作用函数,在完成变换同时有效地把量化与神经网络相结合,在压缩比不变的情况下提高了恢复数据的信噪比.
In this paper, a neural network model that can transform, quantize and encode data is proposed. The discrete staircase quantization function is introduced as the hidden layer neuron function. After the transformation is completed, the quantization is effectively combined with the neural network. When the compression ratio is unchanged The case improves the signal to noise ratio of the recovered data.