自然牙列者叩齿和磨牙时产生的音作频率成分分析及识别

Frepuency Component Analysis and Recognition of Occlusal Sounds Produced By Tap and Grinding of Subjects With Natureal Dentition

选择20名具有完整恒牙列的正常受试者,采集他们在正中(牙合)叩齿和磨牙时所产生的(牙合)音。对所采的(牙合)音建立自回归模型,依据模型系数计算(牙合)音的功率谱,分析其频率成分;并用模型系数建立Bayes'判别函数,对叩牙及磨牙产生的咬合音作识别。结果发现诸如累积功率频率,平均功率频率和谱峰频率等反映频率特征的指标,在二类(牙合)音中呈相似状态;然而Bayes'判别则有95％的识别符合率。结果提示,采用自回归模型的模式识别方法,在临床上诊断(牙合)音中有应用前景。

Twenty normal subjects with intact and permanent dentitions were selected for sampling their occlusal sounds produced by centric tap and grinding. The sampled sounds were processed by autoregressive model, their power frequency spectra were established for analyzing the sound frequency components, and Bayes' discriminant functions built with the AR model coefficients were used to identify two sorts of occlusal sounds, it was found that the indices reflecting frequency features, such as cumulative power frequency, mean power frequency and spectral peak .frquency, were similar in two kinds of occlusal sounds; however, Bayes' iscrimination showed satisfactory recognition conformation rate 95% in total. The results indicate that pattern recognition using AR model has the prospect of identifying and diagnosing occlusal sounds in clinic.