Vous êtes sur la page 1sur 37

US008260778B2

(12) United States Patent


Ghatak
(54) MOOD BASED MUSIC RECOMMENDATION

(10) Patent N0.:


(45) Date of Patent:

US 8,260,778 B2
Sep. 4, 2012
gilfg aser etet all a . ~~~~~~~~~~~~~~~~~~~~~~ .................... .. ~~

METHOD AND SYSTEM

2006/0254411 A1
2007/0133357 A1*

11/2006 Alcalde et al.


6/2007 Korst et al. ..................... .. 369/2

(76)

Inventori

Kallslk Ghatak, Bellevue, WA (Us)

2007/0250319 A1 * 10/2007 Tateishi et al.


2007/0265979 A1* 11/2007

.. 704/251

Hangartner ................... .. 705/59

(*)

Notice:

Subject to any disclaimer, the term of this

2007/0282860 A1

12/2007 Athineos et a1.

patent is extended or adjusted under 35

(Continued)
OTHER PUBLICATIONS

U.S.C. 154(b) by 535 days.


(21)
(22)

Appl' NO: 12/355,742


_

Cont et al., Anticipatory Model of Musical Style Imitation using


Collaborative and Competitive Reinforcement Learning, In Antici

Flled:

Jan 161 2009


_ _ _

patory Behavior in Adaptive Learning Systems, M.V. ButZ et a1.


(Eds): ABiALS 2006, LNAI 4520, pp. 285-306, Springer Verlag,

(65)

Prior Publication Data

Sep 19, 2007

US 2009/0182736 A1

Jul. 16, 2009

(Continued)
Primary Examiner * Angelica Ruiz

Related U-s- Application Data

(60)

Provisional application No. 61/011,391, ?led on Jan. 16, 2008.

(74) Attorney! Agent) 0" Firm *Davis Wright Tremaine LLP; George C. Rondeau, Jr.; Heather M. Colbum

(51)

Int. Cl. G06F 7/00 G06F 17/30

(57)
(2006.01) (2006.01)

ABSTRACT

(52)
(58)

US. Cl. ....................... .. 707/736; 707/748; 707/758


Field of Classi?cation Search ...................... .. None

A method of recommending songs to a listener. The method obtains a set of feature values for each of a plurality of songs. Whenever a playback event is generated for a song, a ?rst
reward and a Mood State Value (MSV) based on a set of

See application ?le for complete search history.


(56) References Cited
U.S. PATENT DOCUMENTS 6,418,424 B1 7/2002 Ho?berg et a1
6,460,036 B1 * 10/2002 HerZ ........................... .. 707/748

parameters are determined. Each feature value associated with the song is associated with the MSV and for each feature value, a second reward is calculated based on the ?rst reward and is assigned to the feature value. A current MSV is deter mined and any songs having a feature value identical to a feature value associated with the current MSV are identi?ed.
For each song identi?ed, a cumulative reward is calculated by

6,748,395 B1
7,196,258 B2

6/2004 Picker et al
3 2007 P att

totaling the second reward assigned to any feature values


' ' ' '

7,376,672 B2 *
7,447,705 B2
2003/0089218 A1 *
2006/0143647 A1 *

5/2008 Weare 707/753


11/2008 Kaiser et 31
5/2003 Gang et al. .................... .. 84/615
6/2006 Bill ..... .. .. 725/10

alssomatedjwlt? the current MSV that are also aSSOCgIIZiWIIE


e song. ma y, one or more songs are recommen e ase
011 thelr Cumulanve rewards

2006/0170945 A1 *

8/2006 Bill ............................ .. 358/l.l3

20 Claims, 20 Drawing Sheets

' 220

2,,
Determine the vaiue of the mood
Parameters 5 Obtain mood State Value

@117
\
Obtain cumulative reward for the song

242

Recommend songs with non ascending order of cumulative

I
224

by adding all reward


values

reward

\ Identify a mood state value in


close proximity to current mood
state value from database

1
I
Identify reward values for all

sort the st of Songs


by cumulative reward

226
A mood state is
. . .

matching feature Values. of the song


associated to the 240

identified mood slate

Yes
2Z8

from the database

S0ng(s) exist with

Identify list of (feature value, reward value) pairs for the identified mood
state from database

positive
cumulative reward?
A song exists in the iist?

1
2* 0 identify list of songs having at least one identified feature value from song collection

232

US 8,260,778 B2
Page 2
U.S. PATENT DOCUMENTS
2008/0021851 2008/0059287 2008/0091717 2008/0114805 2008/0147711 2008/0156173 2008/0257134 2008/0275904
A1 A1
A1 *
A1*

OTHER PUBLICATIONS

1/2008 Alcalde et al. 3/2008 Bonet et al.


4/2008 Garbow et al.
5/2008 Nord ................ ..

Collins, Nick, Reinforcement Learning for Live Musical Agents, Proceedings of ICMC2008, International Computer Music Confer
' 707/l04'l ence, Belfast, 2008.

' 707/ 1041


"" " 707/102 """ " 84/601

Legaspi et al., Positing a Growth-Centric Learning of Empathy


Models in HSI, The 22nd Annual Conference ofthe Japanese Soci ety for Arti?cial Intelligence, 2008, pp. 1-4.
* cited by examiner

A1 *
A1 *

6/2008 Spiegelman et al.


7/2008 Bauer ................... ..

A1
A1*

10/200 8 Oppenheimer
11/2008 Breebaart et al. ..... .. .... .. 707/102

US. Patent

Sep. 4, 2012

Sheet 4 0f 20

US 8,260,778 B2

MwmHEDuGZU
w HMDOH
NM<FLOKD

miE?uwzm
muqw .1532 e52

momlwa qozlw

US. Patent

Sep. 4, 2012

Sheet 7 0f 20

US 8,260,778 B2

:2

QHMIAWDZ
OHAWDMZ

US. Patent

Sep. 4, 2012

Sheet 9 0f 20

US 8,260,778 B2

I\

(OI PIQMGH) 5110s

\r (I DIEMSH) 5110s
(6 plemeg) buog

>I

(3 1019mm) 5110s
(8 pier/x93) buos

>-

(g pxemeg) buog
(L pier/x93) buos
(p {DIED/L8H) buog

9 FIGURE

(9 pier/x93) buog
(g pIQMeH) buog

(buos
E? ;o pJEEMSI eAImQTnumO

o1 [EUOIlIOdOId)

eouejledxg EUIUGlSFI

Vous aimerez peut-être aussi