A sound discriminating device capable of correctly discriminating a cry or other given sounds is provided. When a sound is input from a sound input unit, a feature amount extracting unit extracts a differential value between the amplitude of a fundamental frequency of the input sound and the amplitude of the second harmonic of the fundamental frequency as a feature amount of the input sound. A likelihood calculating unit calculates a likelihood between an acoustic model set for which a feature amount is known and the extracted feature amount. A result output unit determines whether or not the input sound is the given sound based on the result of the likelihood calculation.