Speech translation (ST) is a subject of rapidly increasing interest in the area of speech processing research. This interest is apparent from the increasing tools and corpora for this task. However, the lack of sufficient datasets is still the biggest challenge for under-resourced languages. Specifically, ST requires a large corpus of parallel speech, transcription, and translation text. In this work, we construct a large corpus of the Extraordinary Chambers in the Courts of Cambodia (ECCC), including simultaneous translation from Khmer into English and French. We also address the problem of sentence segmentation of Khmer by conducting a bilingual sentence alignment from English to Khmer with a monotonic assumption. This corpus has approximately 155 hours of speech in length and 1.7M words of text. We also report the baseline results of automatic speech recognition (ASR), machine translation, and ST systems, which show reasonable performance.