中文摘要：

针对冲击地压预警困难这一难题，基于地音监测提出一种新的前兆信息辨识模型及方法。在固定大小的时间窗口内对地音监测信号进行时频域特征提取，得到11个表征冲击地压灾害前兆的多维特征向量，以实际地音监测数据为训练样本，基于SVM理论建立冲击地压多参量前兆信息辨识模型；提出一种新的SVM学习方法，用于解决工程实际应用中的大规模不平衡数据集训练问题，提高SVM分类准确率及速度。利用地音实测数据作为学习样本对支持向量机进行训练，建立相应的前兆辨识模型进行辨识，准确率达到93.87%。实验分析表明，这种方法有效可靠，样本辨识速度快，能够满足在线监测要求，具有工程应用前景。

英文摘要：

A new model and method of identification precursor information was presented based on the monitoring of ground sound to provide an early warning of rockburst. Using the actual monitored data of ground sound as the training samples,eleven multidimensional feature vectors characterizing the rockburst hazard precursors were obtained from the ground sound signals in a fixed time window by the method of time-frequency domain feature extraction. A new support vector machine（SVM） learning method was applied to solve the training problems with imbalanced data sets in the application of large-scale engineering practice and to improve the classification accuracy and the training speed of SVM. Adopting the measured data as the learning samples on SVMs for training and establishing the appropriate precursor identification model,the accuracy reached 93.87%. Experiments show that this method is effective and reliable,and which is capable of fast sample identification meeting the online monitoring requirements.