语言能力是人类区别于动物最本质的能力之一。其中，语音加工是语言认知的核心功能，语音加工的脑机制是语言学及认知心理学关注和研究的重要课题。该领域已有的大部分研究主要关注成人、儿童青少年或婴儿如何加工和处理语音，目前我们对新生儿（人类出生28 天内）的大脑是如何感知语音的尚不清楚。随着神经科学技术的发展，非侵入式的大脑活动测量技术被越来越多的用于考察新生儿的脑机制。研究发现，人类在新生儿时期就已经存在相对完善的语音加工神经系统。例如，感知超音段特征的关键脑区为右侧颞上回，检测音节序列结构的关键脑区为左侧额下回（Broca 区）。本文分别从新生儿对音段和超音段特征的感知、对音节序列结构（包括序列边缘、重复结构、结构分割）的感知，以及对母语和外语感知的差异这三个方面，介绍新生儿对不同语音特征感知的大脑机制，并就此领域的研究发展方向做出了几点探讨。

Language is a unique human trait and a vital means for communication. However, it is unclear how the newborn's brain perceives language and speech sound. It is of great significance to confirm the typical brain mechanism of speech sound perception in neonates. Language with sound carrier is called speech sound, which is produced by human vocal organs. Since newborns are not able to recognize words, neonatal processing of language is generally referred to the processing of speech. With the development of neuroscience, non-invasive neuromeasurement techniques have been increasingly applied to study the brain mechanisms of speech processing in infants and even newborns. This article introduced the brain mechanisms of mother tongue and foreign languages speech perception in newborns from four aspects, i.e., the processing of phonetic structures, phonetic physical attributes, phonetic units and phonetic categories. For the processing of phonetic structures, a number of studies have found that humans are born with the fundamental mechanisms for encoding the order and structure of syllables in speech sequences. It has been shown that newborns have better encodings for the syllables at the edges of sequences than those in the middle. They are able to detect the repetition structures and their positions in syllable sequences. Besides, the statistical learning ability for linguistic sequences is already present at birth. With regard to the processing of physical attributes of speech, significant MMR (mismatch response) are found in response to the change of phonetic physical attributes at the frontal lobe and temporal lobe of newborns. The MMR is the infant equivalent of the mismatch negativity (MMN) in the adult brain. For the processing of phonetic units, the neonatal brain can also elicit the MMR in response to the change of phonetic units. With respect to the processing of phonetic categories, humans are born with a left hemisphere superiority in processing their mother tongue, whereas the right hemisphere is more sensitive to non-native languages. In conclusion, human beings have relatively complete speech processing mechanisms in the neonatal period, which mainly involves temporal lobe and the Broca's area of inferior frontal gyrus. The left inferior frontal gyrus plays an important role in detecting speech structure, while the right superior temporal gyrus is responsible for distinguishing the prosody of the speech.

According to the existing research reviewed above, three questions needing further research and explorations are proposed. Firstly, the polarity of MMR evoked by novel auditory materials needs to be further studied. Longitudinal experiments are suggested to be conducted in the future to reveal the key developmental stage of polarity change of MMR in infants. Secondly, we also suggest that meta-analysis of speech processing effects in different brain regions should be carried out to further reveal the brain mechanism of speech perception in neonates and infants. Lastly, most of the existing studies on the brain mechanism of neonatal speech processing are single time node studies, and do not investigate the obvious changes of the brain stimulated by speech stimulation in the environment of neonatal dependence after birth. In the follow-up study, high spatial resolution technique (fMRI/fNIRS) can be used to further explore the key brain regions of neonatal speech learning.