摘要
矢量量化（ＶｅｃｔｏｒＱｕａｎｔｉｚａｔｉｏｎ）是一种重要的数据编码方法，其关键是码本设计和检索方法。然而，通常的码本聚类生成算法所用的局部信息，不能很好表现信号特征。由Ｋｏｈｏｎｅｎ提出的自组织特征映射（ＳｅｌｆＯｒｇａｎｉｚｉｎｇＦｅａｔｕｒｅＭａｐ，ＳＯＦＭ）具有自组织聚类特性，可直接用于码本设计，并较著名的ＬＢＧ算法有许多良好特性。本文在简略描述自组织特征映射网络特点后，主要论述一种新改进算法并将其运用于心电信号压缩。从压缩结果来看，ＳＯＦＭ法适用于多种心电波形压缩任务；对于基线波动严重的波形，效果可较好；对规则波形如Ｔ１００等却不很理想，Ｓ波处有失真。
Vector quantization (VQ) is an important method for data compression. Now vector quantization using Neural Networks methods received much attention in data compression techniques, especially using Kohonen's selforgamizing feature map (SOFM) in the design of codebook, whose result were super to the LBG algorithm. After discribed the SOFM algorithm and its change forms, this paper presented a new SOFM algorithm and used it to compress ECG data. Experiments showed that this method was better than the traditional SOFM algorithm in ECG data compression, it could be used for various ECG waveforms. The accuracy of reconstruction could be improved if the number of codeword was increased.
出处
《中国生物医学工程学报》
EI
CAS
CSCD
北大核心
1999年第1期97-103,共7页
Chinese Journal of Biomedical Engineering
关键词
心电信号
自组织特征映射
矢量量化
数据压缩
Electrocardiogram (ECG), Self Organizing Feature Map (SOFM), Vecotor Quantization (VQ), Data compression
