The measurement of emotional responses in experimental animals is essential for understanding the neural mechanisms underlying sensory perception. Here, we investigated the utility of 50 kHz ultrasonic vocalizations (USVs) as an indicator of pleasurable emotions related to palatable food consumption in rats. Adult male Sprague-Dawley rat pairs were presented with chocolate, and their USVs were recorded and analyzed using DeepSqueak. The spectrogram of 50 kHz USVs was compared between the anticipatory and consumption phases. Regarding the spectrogram, a decrease in frequency range during the consumption phase compared to the anticipatory phase was observed. A logistic regression-based machine learning model was developed using data from 10 rat pairs to differentiate between anticipatory and consumption phase USVs. Subsequently, 22 rat pairs were divided into groups receiving chocolate or not, revealing that 'consumption phase'-type USVs occurred specifically post-chocolate presentation. These USVs were rarely observed without chocolate. The spectrogram of 'consumption phase'-type USVs predominantly exhibited an inverted U shape or flat shape in the 40 kHz frequency range. Our findings highlight the existence of distinct USV subtypes associated with chocolate consumption, with a machine learning model effectively classifying these subtypes. Future neurobiological experiments should explore whether the 'consumption phase'-type 40 kHz USVs can serve as an index of positive emotion during palatable eating.