Two major auditory models have been proposed to explain the perception of sound pitch. For decades, a significant debate has raged between the temporal theory, which asserts the temporal periodicity of sound as crucial, and the frequency theory (place theory), which is grounded in the sensory organs that detect harmonic components. In this study, we attempted to generate new experimental voices that are commonly heard in everyday life to study voice pitch perception. A source-filter type vocoder, WORLD, was modified to generate a multi-layer dynamic band-noise voice with a spectrum extracted from ordinary speech and driven by white noise. This noise has no periodicity, but a dynamic band noise with a harmonic structure-like spectrum. In the evaluation experiment, we used singing voices as materials. We prepared (1) a simple re-synthesized voice, (2) a pseudo-whispering voice with noise-vocoded spectral envelope, (3) a multi-layered dynamic band-noise voice with the harmonic structure preserved, and (4) a band-noise voice based on transposed singing voices. Next, we conducted auditory evaluation experiments on these experimental voices. Two hundred fifty-six first-year university students listened to the generated experimental voices. They rated the following: a) type of vocalization (a1: whispering voice, a2: singing voice), b) perception of melodiousness, and c) naturalness of the voices. As a result, the multi-layered, dynamic band noise generated was perceived as a "whisper," composed of noise, but it sounded like a "singing voice" with a melody. However, considering the real structure of the human vocal organs, this band noise cannot be produced as a human voice. Nevertheless, since it was recognized as a whisper and the melody could also be evaluated, this sound will be effective in pitch perception research. Furthermore, it can be used as material for creating artistic works, such as a whispering voice that sings.