In this study, the research and implementation of an automatic power quality (PQ) recognition system are presented. This system contains a USB interfaced multichannel data acquisition (DAQ) device and a graphical user interfaced (GUI) application. The DAQ device consists of an analog-to-digital (ADC) converter, field programmable gate array (FPGA) and a USB first in first out (FIFO) buffer interface chip. The application employs Stockwell Transform (ST) technique combined with neural network model to build the classifier. Eight basic and two combined PQ disturbances are determined for the classification. Different from the previous studies, the synthetic signals used for neural network training are modified by adding the harmonics detected in the real signal. This approach is used to increase the classifier accuracy against the real line power signal. Also, ST is simplified by using only the frequencies which are required in the feature extraction step to reduce the processing time. Developed application handles the signal processing, the classification, and the database recording tasks by using multi-threaded programming approach under the mean time of 41 ms. The experimental results show that the proposed power quality disturbance detection system is capable of recognizing and reporting power quality faults effectively within the real-time requirements.