Here are the Roomba robot and Samyukta Ramnath again - this time responding to voice while music is playing in the background.
The Raspberry Pi is mounted on the Roomba Robot, and is communicating with it serially. The microphones are connected via a USB sound card to the RPi, and kept attached to a ruler at a distance of about 8 cm, with the microphones placed approximately across the diameter of the Roomba.
In the bottom left of the screen, a phone is playing an instrumental version of 'Scarborough Fair' by Simon and Garfunkel. Samyukta waits as the Roomba listens to the source of the sound, and starts moving towards the phone.

Then, she starts singing, and the RPi now picks up her voice and starts moving towards her despite the music still playing at a comparable volume level in the opposite direction.
There is a digital low-pass filter on each of the left and right channels of the stereo microphone set. When there is no other sound, the microphone picks up the part of the music that is passed through by the low-pass filter, and moves towards that sound source (the phone playing instrumental music). But when Samyukta starts singing, the voice becomes the dominant signal in that low-frequency range, and it picks up the location of her voice and starts moving towards her.

Stereo Voice Detection and Direction Estimation in Background Noise or Music for Robot Control