Detection of tool wear is vital for the deep-hole drilling, because it can help increasing manufacturing productivity and decreasing tool cost. This paper uses the drilling noise to establish the BTA tool wear condition monitoring system in order to monitor the tool wear condition. After the improved Empirical Mode Decomposition (EMD) method is used to do the modal decomposition for noise signal which has been filtered, the Intrinsic Mode Function (IMF) of signal is obtained. Every IMF is analyzed and detected by the Hilbert-Huang transformation (HHT), and then the energy of marginal spectrum and the changing law of peak value along with the tool wear are extracted. For the relationship between the noise feature vector and tool wear has strong randomness and uncertainty in the process of drilling, so this paper proposes a drill wear state identification method which is based on the genetic support vector machine (GA-SVM). The experimental results show that after dealing the drilling noise signal with HHT, the energy spectrum and the peak spectrum of each frequency band can be extracted as the characteristic vector which can accurately depict the change of drilling system with tool wear. The statistical models of the condition of tool wear established by using GA-SVM can effectively track the trend of tool wear, so as to realize the monitoring of tool wear and tool's life.