2 3 Fig. 2 3-layer structure model ( 3 ) Fig. 1 1 Period for increasing knowledge and familiarity ) 3 ( 2 ) ( 2(a)) 2 6 (

Size: px

Start display at page:

Download "2 3 Fig. 2 3-layer structure model ( 3 ) Fig. 1 1 Period for increasing knowledge and familiarity ) 3 ( 2 ) ( 2(a)) 2 6 ("

あまめはかまや
5 years ago
Views:

1 An Interactive Agent Supporting First Meet Based on Communication Activity Measurement Ryohei Sasama, 1 Tomoharu Yamaguchi, 1 Mutuo Sano, 2 Kenzaburo Miyawaki 2 and Keiji Yamada 1 In this paper, we describe robotic agent which can promote communication on a first meet between persons who don t know each other. We propose a control method based on the communication activity. To establish this method, it is essential to investigate the communication activity and formulate behaviors of our agent, such as a glance, a nod and a suitable timing of those. To evaluate the efficiency of the proposed method, we conducted two kinds of communication experiments of first meeting, 1) mediated by a CG agent with proposed method, and 2) without mediation. As a result, the agent could detect state transitions of communication, control its behavior successfully. The effectiveness of the proposed method was confirmed by the user questionnaire. 1 NEC Corporation, C&C Innovation Research Laboratories 2 Osaka Institute of Technology, Faculty of Information Science and Technology 1. QoL 1) 2) Levinger 3) 1 ( 1(a)) ( 1(b)) 2 ( 1(c)) ( 1(d)) ) 5) 6)7) 1 c 2009 Information Processing Society of Japan

2 2 3 Fig. 2 3-layer structure model ( 3 ) Fig. 1 1 Period for increasing knowledge and familiarity ) 3 ( 2 ) ( 2(a)) 2 6 ( 2(b)) 4 6 ( 1 ) 6 9 ( 2(c)) ) 2 9) Erving Fig states transition model on communication field (1) (2) H1 H2 H1 H2 H1 H2 (3) 2 2 (4) 2 c 2009 Information Processing Society of Japan

3 (5) (6) (a) H1 H2 2 (1) (4) H1 H2 H1 H2 (a) (b) (b) 2 2 (1) H1 H2 H2 H (b) (c) (c) (2) (c) (d) 3 (c) (b) (d) 2 (3) (6) 2 2 (d) (b) ( 4(a)) 2 ( 4(b)) ( 4(c)) ( 4(d)) ( 4(e)) ( 4(f)) ( 4(g)) 4 Fig. 4 Overview of the system c 2009 Information Processing Society of Japan

(a) Greeting (b) Topic offering (c) Focusing on a topic (d) Hearing conversation (1) Leading utterances (2) Gaze lead (3) Gaze distribution (4) Synchronizing nod (5) Sync. to a listener (6) Sync.

4 (a) Greeting (b) Topic offering (c) Focusing on a topic (d) Hearing conversation (1) Leading utterances (2) Gaze lead (3) Gaze distribution (4) Synchronizing nod (5) Sync. to a listener (6) Sync. to a speaker 1 Mediator s behavior of on each state of communication field 5 ( 5(a)) ( 5(b)) 5 2 (1) 2 20% 80% 5 Fig. 5 Example of agent motions 6 Fig. 6 Decision tree for the timing of agent motions Rule1: Rule2: (0.45 ) Rule3: (0.45 ) 2 4 Eval(t) 4 c 2009 Information Processing Society of Japan

5 2 Eval(t) Eval(t) 3 Eval(t) t Eval(t) 3 Eval(t) t X Y N(X, t):x t U(X, t):x t U(X Y, t):x Y t G(X Y, t):x Y t G(X Y, t):x Y t F (X Y, t):x Y t F (X Y, t):x Y t X Y (G(X Y, t) = F (X Y, t) G(X Y, t) = F (X Y, t)) Eval(t, 1) Eval(t, 4) R 2 H1 H2 1. R H1 H2 Eval(t, 1) R H1 R H1 R H2 R H2 Eval(t, 1) = (G(R H1, t)u(r H1, t)u(h1 R, t)) (G(R H2, t)u(r H2, t)u(h2 R, t)) (1) Eval(t, 1) = 1(= T h Eval 1 ) (2) 2. R H1 H2 H2 H1 Eval(t, 2) H1 H2 Eval(t, 2) = (U(H1, t) + U(H2, t))/2 +γ(g(h1 H2, t) + G(H2 H1, t)) +δ(n(h1, t) + N(H2, t)) (3) α β R H2 H1 H1 R(H2 H1) H1 H2 H2 R(H1 H2) R(H2 H1) R(H1 H2) R(H2 H1, t) = U(H1, t) + αn(h1, t) + βg(h1 H2, t) R(H1 H2, t) = U(H2, t) + αn(h2, t) + βg(h1 H2, t) (4) α β R(H2 H1) R(H1 H2) R R(H2 H1) R(H1 H2) T h React Eval(t, 2) T h Eval 2 R H1 H2 R(H2 H1) T h React R(H1 H2) T h React Eval(t, 2) T h Eval 2 (5) 3. R Eval(t, 3) R H1 H2 H1 H2 Eval(t, 3) = (U(H1, t) + U(H2, t))/2 +γg(h1 H2, t) +δ(n(h1, t) + N(H2, t)) (6) α β γ δ γ δ Eval(t, 3) T h Eval 3 R H1 H2 5 c 2009 Information Processing Society of Japan

6 Eval(t, 3) = T h Eval 3 (7) 4. R H1 H2 Eval(t, 4) R H1 H2 ((6) ) Eval(t, 4) Eval(t, 4) = Eval(t, 3) (8) T h Eval 4 > T h Eval 3 > T h Eval 2 > T h Eval 1 (9) Eval(t, 4) T h Eval 4 (10) 4. Fig. 7 7 ( ) Detection of gazing direction 2 ( ) 2 ( ) ( 7(a,b)) 7(c,d,e,f) ( 7(g)) 8 ( 8(a)) ( 8(b)) 9 ( 9(a)) 2 ( 9(b,c)) Fig. 8 8 Acceleration sensor and microphone Eval(t) c 2009 Information Processing Society of Japan

(a) Agent condition Statistical method Questionnaire item (b) Did you come to know your partner well? (c) Did you feel your partner favorable? (d) Was the conversation enjoyable? Existing Average 4.

7 (a) Agent condition Statistical method Questionnaire item (b) Did you come to know your partner well? (c) Did you feel your partner favorable? (d) Was the conversation enjoyable? Existing Average Median Standard deviation Nothing Average Median Standard deviation Table 2 Statistical summary of questionnaire ( 1 ) : ( 1 ) ( 1 ) 1 : Fig. 9 9 A scene of the experiment 10 A 1 B 5 2 C 1 2 A B C A : ( 1 ) 1 : : ( 2 ) B 2 : 1 : 2 : ( 5 ) ( 2 ) ( 2 ) :( 2 ) 1 : 2 : 10 C 2 : ( 2 ) 1 : ( 2 ) 2 : 5. ( 2(a)) (Existing) (Nothing) c 2009 Information Processing Society of Japan

8 Fig (Eval(t)) Results of communication activity(eval(t)) and state transition on communication field ( 2(b)) ( 2(c)) ( 2(d)) t 2 5 % sec 300sec 1),,, : 3,, pp , (1998). 2) : 19, 2, (2007). 3) Levinger G., Snoek D. J.: Attraction in relationship: A new look at interpersonal attraction, General. Learning Press, (1972). 4),,, :,, Vol.42, No.6, pp , (2001). 5),,,, : Social Context-aware Information System:, 2006,, Vol.2006, Num.4, pp , (2006). 6) Kenzaburo M., Mutsuo S., Yoshinari T., Tatsuya H., Kentaro, M., Masao K., Ryohei S., Hirohiko I., Tomoharu Y., Keiji Y.: Adaptive Embodied Entrainment Control based on Communication Activity Measurement A Challenge for A Robotic Introducer Agent, Proceedings of the 4th International Conference on Autonomous Robots and Agents (ICARA2009), pp , (2009). 7),,,,,,,,, :, IECE Technical Report HCS , pp.49 54, (2009). 8) : 9, 1 1 4, (1997). 9) Erving G.: Forms of Talk, University of Pennsylvania Press, (1981). 8 c 2009 Information Processing Society of Japan

& Vol.5 No (Oct. 2015) TV 1,2,a) , Augmented TV TV AR Augmented Reality 3DCG TV Estimation of TV Screen Position and Ro

& Vol.5 No (Oct. 2015) TV 1,2,a) , Augmented TV TV AR Augmented Reality 3DCG TV Estimation of TV Screen Position and Ro TV 1,2,a) 1 2 2015 1 26, 2015 5 21 Augmented TV TV AR Augmented Reality 3DCG TV Estimation of TV Screen Position and Rotation Using Mobile Device Hiroyuki Kawakita 1,2,a) Toshio Nakagawa 1 Makoto Sato