IEEEXplore - [exploratory DSP] Alex (Sandy) Pentland Social...

Info iconThis preview shows pages 1–2. Sign up to view the full content.

View Full Document Right Arrow Icon
IEEE SIGNAL PROCESSING MAGAZINE [ 108 ] JULY 2007 [ exploratory DSP ] Alex (Sandy) Pentland 1053-5888/07/$25.00©2007IEEE F ace-to-face communication conveys social context as well as words. It is this social signaling that allows new information to be smoothly integrated into a shared, group-wide understanding. Social signaling includes signals of inter- est, determination, friendliness, bore- dom, and other “attitudes” toward a social situation. Psychologists speculate that social signaling may have evolved as a way to establish hierarchy and group cohesion [1]–[3] because social signaling functions as a subconscious discussion about relationships, resources, risks, and rewards. In many situations the nonlin- guistic signals that serve as the basis for this social discussion are just as impor- tant as conscious content for determin- ing human behavior [1]–[5]. In what follows we discuss challenges in exploratory processing of social sig- nals and tools that allow us to predict human behavior and sometimes exceed even expert human capabilities. These tools potentially permit computer and communications systems to support social and organizational roles instead of viewing the individual as an isolated entity. Example applications include automatically patching people into socially important conversations, insti- gating conversations among people in order to build a more solid social net- work, and reinforcing family ties. CHALLENGES IN SOCIAL SIGNAL PROCESSING To use social signals in computer and communications systems we must first define them in terms of signal properties and then develop automatic detection and measurement methods—just as the speech recognition community has done for detection and classification of words in speech. However, human gestures and variations in vocal prosody have proven very difficult to measure outside of care- fully controlled conditions; and even when they can be measured, their inte- gration into a communications system has proven quite challenging. To address these difficulties we have been working to develop digital signal processing tools that measure social sig- nals by analyzing the statistical proper- ties of the speaker’s tone of voice, facial movement, and gesture [1]. These statis- tical properties are based on relatively easy-to-measure properties of the signal such as voicing segment duration and gesture optical flow. Because our meth- ods employ statistical “texture” descrip- tions instead of attempting to perform detailed recognition of gestures and their features, they are much more robust to noise and distortion. We have found that by using “texture” measures of prosody and gesture, we can reliably measure social signaling and can begin to build communication tools that are aware of the social context. Our tools for meas- urement of nonlinguistic social signals have proven to be particularly powerful for prediction of human behavior. SOCIAL INTERACTION AND
Background image of page 1

Info iconThis preview has intentionally blurred sections. Sign up to view the full version.

View Full DocumentRight Arrow Icon
Image of page 2
This is the end of the preview. Sign up to access the rest of the document.

Page1 / 4

IEEEXplore - [exploratory DSP] Alex (Sandy) Pentland Social...

This preview shows document pages 1 - 2. Sign up to view the full document.

View Full Document Right Arrow Icon
Ask a homework question - tutors are online