This work integrated several features extraction techniques for a raw sensor data from a wearable assistant consists of three accelerometers sensors for patients suffered from Parkinson's disease (PD) and Freezing of Gait (FOG) symptom during their movement. We considered three types of transformation, namely the one-dimensional Discrete Wavelet Transform (1D DWT), the two-dimensional Discrete Wavelet Transform (2D DWT), and the Fast Fourier Transform (FFT). The extracted features from these transformations were applied to machine learning methods, such as Artificial Neural Network (ANN) to detect FOG. The proposed hybrid system integrates the extracted features from 1D DWT and FFT concluding a total of fifteen extracted features. These hybrid features are then used during the classification process using the ANN that established accuracy of 96.28% for PD detection. The results established that the hybrid 1D DWT-FFT features has the ability to build a robust classification model to detect the FOG accurately.