Infants often pay special attention to speech sounds, and they appear to detect key features of these sounds. To investigate the neural foundation of speech perception in infants, we measured cortical activation using near-infrared spectroscopy. We presented the following three types of auditory stimuli while 3-month-old infants watched a silent movie: (1) normal speech sounds; (2) sine wave speech (SWS) sounds consisting of three sine waves that tracked the first, second, and third formants of speech sounds; and (3) synthesized tones composed of three pure tones. Statistical analyses of oxygenated hemoglobin (oxy-Hb) signals revealed significant activation in the left and right auditory areas in all conditions. Direct comparisons of oxy-Hb signal changes between SWS and synthesized tones showed significant differences in the left frontal and temporal regions. Furthermore, comparisons of oxy-Hb signal changes between speech sounds and SWS exhibited significant differences in a left posterior temporal region. These results demonstrated that functional differentiation occurs in the left temporal cortex while infants perceive different types of auditory information. Coactivation of the left temporal and frontal regions by speech sounds suggests the initial formation of a left fronto-temporal network related to infant speech processing. Clarification of the functional role of this left-lateralized network will help understand the speech code.