Voices are prominent stimuli in the auditory environment due to the wide variety of ecologically relevant information they code in their acoustic structure, including vocalizer identity cues. Acoustics of voice-based individual recognition is well-described in intraspecific-, but unknown in interspecific contexts. Dogs, for whom the identification of humans became crucial during domestication, are suitable subjects of such investigations. In a behavioural test, dogs had to find their hiding owner based on vocal cues in two-way choice tasks. Stimuli were pre-recorded neutral speech sentences from the owner and a set of control persons played through loudspeakers (placed near the hiding persons) from behind two opaque screens. The effect of speakers’ acoustic distance along a set of dimensions on choosing success, choosing latency and looking time was investigated. Dogs chose their owner’s voice significantly more often than control persons’ voices (82%) and acoustic distance of speakers did not affect their choosing success, which suggests that dogs can confidently identify their owner’s voice. Fundamental frequency-related (F0 mean, gender difference of speakers) and noisiness parameters (entropy, jitter) affected either choosing latency or looking time which, however, indicate that the shorter the acoustic distance between speakers, the harder the decision. According to these results, interspecific voice recognition in dogs is based on similar acoustic parameters as intraspecific voice recognition in both humans and in various canine species. This novel experimental design is applicable to further explore dogs’ interspecific voice recognition ability by artificially modifying the voices.