Abstract: This paper applies the multi-sensor information fusion technology to Coal Mine Safety Monitoring.Combined with the actual environmental of coal mine,a grading architecture was proposed.Bayes estimation was used in data-level,through set soft and hard threshold to avoid sensor-data redundancy.Based on the fuzzy set theory,the result of data-level fusion was fuzzed in decision fusion.Accurate estimation of Coal Mine Safety obtained by means of expert knowledge database and decision rules.This architecture not only ensures the accuracy of the result,but also consideration to the system load.