Early fault diagnosis is imperative for the proper functioning of rotating machines. It can reduce economic losses in the industry due to unexpected failures. Existing fault analysis methods are either expensive or demand expertise for the installation of the sensors. This article proposes a novel method for the detection of bearing faults and imbalance in induction motors using an antenna as the sensor, which is noninvasive and cost-efficient. Time-varying S11 is measured using an omnidirectional antenna, and it is seen that the spectrogram of S11 shows unique characteristics for different fault conditions. The experimental setup has analytically evaluated the vibration frequencies due to fault and validated the characteristic fault frequency by applying FFT analysis on the captured S11 data. This article has evaluated the average power content of the detected signals at normal and different fault conditions. A deep learning model is used to classify the faults based on the reflection coefficient ( S11). It is found that classification accuracy of 98.2% is achieved using both magnitude and phase of S11, 96% using the magnitude of S11 and 92.1% using the phase of S11. The classification accuracy for different operating frequencies, antenna location, and time windows are also investigated.