早期高血压激光、电刺激疗法的信号自动识别研究

Automatic identification of the signal for early hypertension treatment with laser and electrical stimulations

目的:针对早期高血压激光、电刺激治疗中需要多种模式刺激信号以解决患者个体化差异以及适应性的问题,提出了一种刺激信号的自动识别方法,为临床选择有效的治疗信号提供了方便、快捷的筛选手段。方法:用已被临床试验证实有疗效的数字音频信号作为原始信号,提取该类信号的MFCC特征参数建立HMM模型库;将实验用的信号分组,有效信号识别后定出判别标准;对识别信号提取特征参数,利用维特比(Viterbi)算法计算识别信号与模型匹配程度,将识别信号与模板库中的信号自动匹配,根据匹配结果判断识别信号在模板库中所属的类别。结果和结论:待识别信号采用HMM模型能有效地分类识别,并且可从大量的待识别信号中准确、快捷地分拣出识别信号与原始信号输出概率对数的相对值小的信号,待临床验证后用作早期高血压激光、电刺激疗法的信号源。

Objective Based on the situation that simulation signals of different models are needed to solve the problems of individual differences and adaptability in early hypertension treatment with laser and electrical simulations,to propose a method of automatically identifying the simulation signals for providing a convenient and efficient means to choose effective treatment signals in clinic.Methods The digital audio signals which have been clinically proven to be effective were taken as the original signals,and the Mel frequency cepstral coefficients of this kind of signal were extracted to establish the HMM model library.After grouping the experimental signals and identifying the effective signals,the criterion of automatic identification was determined.Then the characteristic parameters of the test signals were extracted,and the matching probability of the test signals and model was calculated by Viterbi algorithm,and the test signals were automatically matched with the signals in the template library.Finally,according to the matching results,the test signals were classified into different categories in template library.Results and Conclusion HMM model could be used to classified test signals effectively.The signals with small relative value of the output probability logarithm between test signals and original signals could be sorted out accurately and effectively from a large number of test signals.After the clinical validation,the selected signals can be used as the signal source for early hypertension treatment with laser and electrical simulations.