This paper presents an improved technique for detection of mass unbalance in rotor of an induction motor by the analysis of transient stator current during starting period using both discrete wavelet transform (DWT) and continuous wavelet transform(CWT). The unbalanced magnetic pull due to centrifugal force developed for mass unbalance in rotor produces excessive vibration in the rotor as well as in the stator, resulting changes in air gap flux distribution which induces voltage in the rotor and in the stator and generation of unique signature pattern in the stator current. During starting, unbalanced magnetic pull is very high, the rotor is pulled across the whole air gap which increases the amplitudes level of the harmonics already present and generation of new harmonics due to this fault in the stator current. In the present paper, the starting current at no load is captured and analysed for diagnosing the motor fault. As motor current signature (MCS) significantly depends on loading condition of the induction motor during steady state, the detection of faults using MCS at no load or light load by FFT technique is difficult. This has been overcome by this proposed improved technique as motor current does not depend on load and is very high even at no load during starting. This technique has been developed and implemented in the laboratory using machine fault simulator.