That can be done using the "Stop command if target window focus is lost" option, within those individual commands.
You'll need to set a target, either at the profile level (which is preferred, so that if that target ever needs to be changed, you only need to do so once), or for each individual command at the command level.
Given that you're indicating that you wouldn't be at your computer, do note that a (USB) headset is the recommended input device for speech recognition.
Using a (desktop) microphone that's located at your computer while you're elsewhere could lead to low recognition accuracy, and especially with the very short command phrases in your example, lead to unintended recognition.
If you're playing music through speakers, that would exacerbate the issue, as would having conversations while the speech recognition system is listening (it is designed to recognize speech, so that's what it'll attempt to do; it can't know whether your actually intending to issue a command).
Some type of wireless remote may be more suitable to your intended application. A wireless presentation remote (I.E. for slideshows/PowerPoint presentations) could be suitable, taking into account the size of the device, and the distance from your computer.
You would need to either configure your music player to respond to the specific keypresses the remote produces, or use an application like VoiceAttack to essentially re-map them to the keyboard shortcuts your music player is expecting.