Meter is one of the core features of music perception. It is the cognitive grouping of regular sound sequences, typically for every 2, 3, or 4 beats. Previous studies have suggested that one can not only passively perceive the meter from acoustic cues such as loudness, pitch, and duration of sound elements, but also actively perceive it by paying attention to isochronous sound events without any acoustic cues. Studying the interaction of top-down and bottom-up processing in meter perception leads to understanding the cognitive system’s ability to perceive the entire structure of music. The present study aimed to demonstrate that meter perception requires the top-down process (which maintains and switches attention between cues) as well as the bottom-up process for discriminating acoustic cues. We created a “biphasic” sound stimulus, which consists of successive tone sequences designed to provide cues for both the triple and quadruple meters in different sound attributes, frequency, and duration, and measured how participants perceived meters from the stimulus in a five-point scale (ranged from “strongly triple” to “strongly quadruple”). Participants were asked to focus on differences in frequency and duration. We found that well-trained participants perceived different meters by switching their attention to specific cues, while untrained participants did not. This result provides evidence for the idea that meter perception involves the interaction between top-down and bottom-up processes, which training can facilitate.