Advanced Auditory Modeling for Acoustic Analysis

The human auditory system out-performs all current machine-based systems for analyzing and interpreting real world acoustic environments. Central to that human performance is the method of auditory scene analysis used by listeners and the mechanisms that allow creation of auditory objects. BioMimetic Systems has implemented real-time biomimetic algorithms in embedded hardware based on the ‘what’ and ‘where’ auditory pathways which are integral components for the formation of auditory objects in the brain. These algorithms provide biologically inspired feature sets representing spike activity from the auditory periphery to various brain centers including the Superior Olive (SO) and Inferior Colliculus (IC), representing critical points in the ’what’ and ‘where’ processing streams. Under this proposal we will accomplish two basic tasks. The first is to unify the features created by these auditory models to create auditory objects as a first step in creating an auditory scene analysis system. The second is to develop models of attentional mechanisms that can be used to modulate the formation of these auditory objects that include neuronal oscillations as well as predictive and competitive mechanisms. The hardware/software system will be tested against simple speech based attention tasks and performance compared to psychophysical data. BENEFIT: The proposed system of biomimetic algorithms and hardware with intrinsic attentional mechanisms will yield significantly increased performance for detection, identification, and localization tasks for acoustic targets. These gains will come in part from the greatly improved signal to noise ratio (SNR) the attentional mechanisms can provide through segregation of component sound sources and enabling analysis on these individual components. There are numerous applications of this technology for military senors in difficult and cluttered battlespace environments, particularly for urban and mountainous terrain with strong reverberation, as well as commercial speech to text systems, cell phone systems, and for advanced robotics.