the-midnight-paper

Whole-home gesture recognition using wireless signals

Authors: Qifan Pu, Sidhant Gupta, Shyamnath Gollakota, and Shwetak Patel

Conference: Mobicomp 2013, Miami, FL USA

Keywords: Gesture Recognition, Wireless Sensing

What is the high-level research question

This paper discusses the use of wireless systems to enable gesture recognition.

Such a system would enable whole-home gesture recognition using few wireless sources without the use of sensors on the human body. Also avoids the use of cameras. Thus the system is scalable and minimizes costs.

Specific Research Question and Hypothesis

This paper addresses the following questions:

Key Findings

Data Collection Methodologies

Received Wi-Fi signals are transformed to a narrow-band pulse (BW of a few Hz). The receiver tracks this frequency to detect small Doppler shifts associated with gestures movements. The receiver also tracks the frequency corresponding to the maximum energy and corrects for the residual frequency offset. The receiver then looks for segments corresponding to the Doppler shifts and clusters these segments into gestures. Gestures are classified by matching the pattern of positive and negative Doppler shifts.

Data Analysis Methodologies

To evaluate how well WiSee can detect the presence of a gesture, the Doppler SNR (the ratio between the average energy in the non-DC frequencies in the profile, with and without the gesture) is computed from the frequency-time Doppler profile. The average Doppler SNR was computed at each location by having each user repeat the gesture ten times. The experiments were carried out in the following scenarios:

Did the paper draw convincing conclusions using the methodology?

The paper proposes a novel and efficient method to enable gesture recognition without the use of expensive and sophisticated technology. However, we do believe that the paper suffers from the following limitations:

Describe a few ways in which the data collection and analysis methodology can be improved to better answer the research questions.

The authors should try to address the gesture-recognition problem using the more widely-used 2.4GHz band so as to make the technology more accessible to everyday users.