In the last 20 years, we have seen the nourishing environment for the development of music software using a corpus of audio data expanding significantly, namely that synthesis techniques producing electronic sounds, and supportive tools for creative activities are the driving forces to the growth. Some software produces a sequence of sounds by means of synthesizing a chunk of source audio data retrieved from an audio database according to a rule. Since the matching of sources is processed according to their descriptive features extracted by FFT analysis, the quality of the result is significantly influenced by the outcomes of the Audio Analysis, Segmentation, and Decomposition. Also, the synthesis process often requires a considerable amount of sample data and this can become an obstacle to establish easy, inexpensive, and user-friendly applications on various kinds of devices. Therefore, it is crucial to consider how to treat the data and construct an efficient database for the synthesis. We aim to apply corpusbased synthesis techniques to develop a Computer Assisted Composition program, and to investigate the actual application of the program on ensemble pieces. The goal of this research is to apply the program to the instrumental music composition, refine its function, and search new avenues for innovative compositional method.