Failure alarm system and method of failure alarming thereof

1. A failure alarm system adapted for monitoring an instrument, comprising:

a BUS transmitting data and control logics inside the failure alarm system;

a recording device connected to the BUS for recording operating sounds of the instrument;

a voice processing unit connected to the BUS for converting the operating sounds into a voice eigenvector;

a storage unit connected to the BUS for storing the voice eigenvector at a normal working state of the instrument as historical data and also storing the voice eigenvector at an abnormal working state of the instrument therein;

a comparing unit connected to the BUS for comparing a new voice eigenvector with the previous voice eigenvector saved as the historical data to judge whether the instrument is at the abnormal working state; and

a warning unit connected to the BUS for warning monitoring personnel when the abnormal state is detected.

2. The failure alarm system as claimed in claim 1 , wherein the recording device is a directional microphone.

3. The failure alarm system as claimed in claim 1 , wherein the warning unit is a monitor screen.

4. A method of failure alarming of a failure alarm system, comprising:

recording operating sounds of an instrument with a recording device in a controlled situation where the instrument is ensured to work in a normal working state periodically, converting the operating sounds recorded in the controlled situation into a voice eigenvector using a voice processing unit and storing the voice eigenvector into a storage unit as historical data;

keeping on recording the operating sounds of the instrument in actual operation with the recording device;

converting the operating sounds recorded in actual operation into a new voice eigenvector using the voice processing unit;

comparing the new voice eigenvector with the previous voice eigenvector saved as the historical data to judge whether the instrument is at an abnormal working state; and

warning monitoring personnel when the compared result shows the abnormal working state occurs, and storing the new voice eigenvector of the instrument at the abnormal working state into the storage unit.

5. The method as claimed in claim 4 , wherein the steps for converting the operating sounds into the voice eigenvector comprising:

finding an end-point of the operating sounds in a wave form to take voice data of voice regions;

breaking the voice data into several frames of which each goes through a very short time period and contains a unit block of sound;

applying a pre-emphasis filter to each frame to increase a resonance peak value within a particular frequency band with respect to magnitudes of other frequencies in order to improve an overall signal-to-noise ratio;

applying a window function to each frame;

converting the windowed frames from a time domain into a frequency domain with Fast Fourier Transform; and

using a Mel-filter band and the Discrete Cosine Transform to determine Mel-scale frequency cepstral coefficients and the voice eigenvector.

6. The method as claimed in claim 5 , wherein the window function is a hamming window function to improve continuity of data at two ends of the frame.

7. The method as claimed in claim 4 , wherein the method for comparing the voice eigenvector and the historical data uses Dynamic Time Warping to determine the degree of difference therebetween.

8. The method as claimed in claim 4 , further comprising the step of stopping the instrument automatically when the instrument is at the abnormal working state and further comparing the voice eigenvector at the abnormal working state of the instrument with the historical data at the normal working state of the instrument to help the monitoring personnel to find out a fault reason of the abnormal working state.

9. The method as claimed in claim 4 , wherein when the compared result shows the instrument works normally, the new voice eigenvector is stored into the storage unit.