In gear condition monitoring, it is often impractical to measure the vibrations directly at or close to their sources. It is common practice to measure the vibration at a location away from the source due to the limitation of accessing to the machine to be monitored. The vibration measured in this way inevitably has high distortions from the vibrations due to the effect of the attenuation of signal paths and the interference from other sources. Suppressing these distortions to obtain useful information is a key issue for remote measurement based condition monitoring. This paper develops a new feature designed for remote gear damage diagnosis based on advanced signal processing. Time synchronous average (TSA) is firstly used to suppress the random noise and interferences. Then Continuous Wavelet Transformation (CWT) of TSA signals is obtained to enhance the fault characteristics. Finally wavelet peak factors and kurtosis are developed as a set of complementary features to classify different faults. The diagnosis results show that these two features can be used to detect and indicate the severity of the gear damage effectively even if vibration signals from a remote motor casing.