Hi Zia,
If you are using an object of the kWaveTransducer class, it is only possible to have a single linear array. However, if you define sensor.mask to be a binary matrix or cuboid corners, you can define an arbitrary number of sensor points in any configuration.
You can assign any signal as your source (e.g., a recorded voice). You will need to assign this to source.p. Keep in mind that you need to make sure the sampling rate used in the recording matches the sampling rate (i.e., 1/kgrid.dt) used in the simulation.
You might find it helpful to work through the examples included in the toolbox from the beginning, as well as reading the k-Wave manual. They contain lots of helpful information relevant to your questions.
Good luck,
Brad.