Well I'm kinda new to electronics and stuff but I would try: a microphone, an ADC converter, a non-volatile memory chip and a microcontroller (perhaps a PIC since its popular around these parts). I would use the ADC converter to turn the voice audio into its digital representation store that digital representation into the non-volatile memory, then when you give the voice command next time the microcontroller retrieves the data stored in the memory chip and compares it to the voice audio that was last received and if it matches, perform a specific task. There are of course other thing you will need to add/set-up - like a button so that the device doesn't continually read voice audio.
NOTE: I'M NOT SURE IF THIS WOULD WORK, THIS IS JUST MY IDEA