We have developed a genuine method for sound and speech recognition, which can be described as having three parts, a pre-processing stage, a sample classifier stage and a time-series analysis stage. The preprocessing stage is a modification of the well-known SPHINX sound processing, while the classification can be done by any multi-class classifiers. We have used a modification of the FuzzyART algorithm made suitable for online-learning mode. At the end we have created a hierarchical time-series analyzer, which can also act as a content-based index in multimedia database applications.

Technologies used:

• ASPHINX
• FuzzyART algorithm