This paper presents a direct torque control technique for Brushless DC motors with non-sinusoidal back electromotive force. Direct torque control has some benefits such as faster torque response and reduced torque ripple for driving the Brushless DC motors. The Sliding mode observer, which is robust to parameter uncertainties, can be used to estimate the back-EMF and generate the torque. It is employed to estimate the non-sinusoidal back-EMF waveform in a Brushless DC motor using only the measurements of the stator currents. A modified version of Non-Dominated Sorting Genetic Algorithm (NSGA-II) is used as an effective optimization tool for regulating the PID controller parameters of speed controller and selection of four gains of sliding mode observer. The relation between PID controller and the sliding mode is also satisfied using NSGA-II. The effectiveness of proposed system has been validated by simulation results.