Accent is a way in which pronunciation of words differ from one another which depends on an individual or groups' location, social class, or birth area. A single language spoken by different regions will have different accents; this fact created a problem which resulted miscommunication due to different accents. This study aims to lessen, if not remove, the hindrance between native speakers understanding non-native speakers in real-time with the utilization of the speak recognition using Hidden Markov Model. The study was used t-test statistical analysis method to check device accuracy. The materials used were microphone, speaker, amplifier and Altera DE0-nano FPGA. The voice input was obtained from the integrated microphone in the laptop and input voice signal was converted into digital signal by the built-in analogue to the digital audio signal converter. The converted signal was stored in Altera DE 0 FPGA for the signal analysis. The analysed and corrected digital signal was converted into analogue signal by A/D converter audio converter to be output through speaker. Meanwhile, the hardware description language (HDL) module was used to synchronize converted audio signal into digital signal through Fast Fourier Transform (FFT) which activated voice learn database, the MATLAB resampled both signal to common scale to search for similar signal using Hidden Markov Model. The system would search most similar signal from the voice learn database using Hidden Markov Model in the native accent database. The system was displayed the accent different between input signal and signal from the native accent database and output signal from native accent database. The results showed all words had t values less than T-distribution value which was 2.145, as well as had - values greater than significant level which was 0,05. These results proved accent correction implementation was successful in the non-native speakers' accent to native speaker accent conversion.