Blackfoot is an endangered Native American language. It is important to document this language and with it, the Blackfoot culture. In this paper, an effective subspace-based concept mining framework (SCM) is used to help Blackfoot language analysis via audio classification. The core of SCM is a subspace based modeling, classification and decision fusion mechanism which is applied to the audio features for pattern discovery. It adaptively selects non-consecutive principal dimensions to form an accurate modeling of a representative subspace based on statistical information analysis and refines the training data set via self-learning. After the classification process, a decision fusion process is applied to traverse the results from individual classifiers and to boost classification accuracy.