摘要
针对混合路面识别问题,提出一种基于概率回声状态网络(PESN)的辨识方法。首先将M分类问题分解成M(M-1)/2个二分类问题,利用Sigmoid函数将回声状态网络(ESN)的数值输出映射为概率输出,然后采用成对耦合法融合ESN的二分类概率,最终提取待识别路面功率谱特征,求解其分属于不同等级路面的概率。试验结果表明,PESN能够有效地识别出混合路面成分及相应比重,从而证明了PESN在混合路面识别中的有效性。
The Probabilistic Echo State Network (PESN) based identification algorithm was proposed for mixed road identification. Firstly M-classified problems were divided into M (M-1 )/2 dichotomy problems; then Sigmoid function was used to transform Echo State Network (ESN) numerical output into probabilistic output; afterwards all the ESN dichotomy probability was fused by pairwise coupling. Finally the power spectrum characteristic of the road under recognizing was extracted, which was utilized to solve the probability belonging to different road levels. The experiment result shows that PESN can effectively identify the mixed road composition and the corresponding proportion, which proves the validity of the proposed PESN in mixed road recognition.
出处
《汽车技术》
CSCD
北大核心
2017年第10期49-53,共5页
Automobile Technology