Active Auditory Processing

Participants: Tomas Figliolia, Andreas Andreou

The approach employed to process and analyze the data (kitchen actions) consists of two stages: the segmentation stage, which obtains a set of features or events from each micro-Doppler signature (Zhang 2007); and the categorization stage, which calculates representative prototypes from these events and uses them to decide what action is being per- formed given a set of incoming test events; this is described in detail in the paper by Dura et. al. (Dura 2011). The different steps are represented schematically below in Figure 4. Even though this approach yielded good results for the classification of kitchen actions, the uDoppler sensors were not fully incorporated in the final demo due to integration and interface issues. It is anticipated that the active auditory processing will incorporated in the overall architecture in forthcoming collaborative effort between the JHU and the UMD group.


(Zhang 2007) Zhaonian Zhang, P. Pouliquen, A. Waxman, and A. G. Andreou, “Acoustic {micro-Doppler} gait signatures of humans and animals,” 41st Annual Conference on Information Sciences and Systems (CISS 2007), pp. 627–630, 2007.

(Dura 2011) S. Dura-Bernal et al., “Human action categorization using ultrasound micro-Doppler signatures,” Second Int. Workshop on Human Behavior Understanding (HBU 2011)(submitted), pp. 1–9, Nov. 2011.