## A Probabilistic Expert System for Automatic Musical Accompaniment (1999)

Venue: | Journal of Computational and Graphical Statistics |

Citations: | 30 - 5 self |

### BibTeX

@ARTICLE{Raphael99aprobabilistic,

author = {Christopher Raphael},

title = {A Probabilistic Expert System for Automatic Musical Accompaniment},

journal = {Journal of Computational and Graphical Statistics},

year = {1999},

volume = {10},

pages = {487--512}

}

### Years of Citing Articles

### OpenURL

### Abstract

A methodology is presented that allows a computer to play the role of musical accompanist in a non-improvised musical composition for soloist and accompaniment. The modeling of the accompaniment incorporates a number of distinct knowledge sources including timing information extracted in real-time from the soloist's acoustic signal, an understanding of the soloist's interpretation learned from rehearsals, and prior knowledge that guides the accompaniment toward musically plausible renditions. The solo and accompaniment parts are represented collectively as a large number of Gaussian random variables with a specified conditional independence structure --- a Bayesian Belief Network. Within this framework a principled and computationally feasible method for generating real-time accompaniment is presented that incorporates the relevant knowledge sources. The EM algorithm is used to adapt the accompaniment to the soloist's interpretation through a series of rehearsals. A demonstration is provided from J.S. Bach's Cantata 12.

### Citations

8526 | Maximum Likelihood from Incomplete Data via the EM Algorithm
- Dempster, Laird, et al.
- 1977
(Show Context)
Citation Context ...O ), the learnable vectorsL = ( 1 L ; : : : ;sJ L ), and the actual observation vector ff = (ff 1 ; : : : ; ff J ). Thus our observed information is summarized by xO = ff. The well-known EM algorithm =-=[12]-=- starts with an initial value for the parameter, ` 0 , and produces a sequence of values ` 1 ; ` 2 : : : that is known to increase the likelihood at every iteration provided that a local maximum has n... |

1131 | Graphical Models
- Lauritzen
- 1996
(Show Context)
Citation Context ...Y fl2\Gamma f fl (x fl jx pa(fl) ) (9) where the parents of x fl , pa(x fl ), are nodes in \Gamma that have edges leading to fl and f fl is the conditional density for x fl given its parents x pa(fl) =-=[7]-=-. If A; B; S ae \Gamma, we use the notation xA ? xB jx S to mean that xA and xB are conditionally independent given x S . Eqn. 9 can be shown [7] under fairly general conditions to be equivalent to th... |

949 |
An introduction to Bayesian network
- Jensen
- 1995
(Show Context)
Citation Context ...ce we have described a means of computing posterior distributions on model variables, given the observation of other model variables. The literature on Bayesian Belief Networks, for example [8], [7], =-=[9]-=-, addresses this problem thoroughly and we utilize that theory here. In particular we incorporate Lauritzen's specialization of the BBN theory to the case of Gaussian distributions [10]. A brief accou... |

224 |
The EM algorithm for graphical association models with missing data
- Lauritzen
- 1995
(Show Context)
Citation Context ...tomatic method for learning the parameters that model this interaction using the EM algorithm. The basic idea we use for training our BBN is thoroughly developed for the case of discrete variables in =-=[11]-=-, and is briefly sketched for the joint Gaussian case which applies to our situation. The probabilistic model for the entire collection of random variables, x \Gamma , presented in Section 2 is descri... |

194 |
Bayesian analysis in expert systems
- Spiegelhalter, Dawid, et al.
- 1993
(Show Context)
Citation Context ...omplete once we have described a means of computing posterior distributions on model variables, given the observation of other model variables. The literature on Bayesian Belief Networks, for example =-=[8]-=-, [7], [9], addresses this problem thoroughly and we utilize that theory here. In particular we incorporate Lauritzen's specialization of the BBN theory to the case of Gaussian distributions [10]. A b... |

142 | Propagation of probabilities, means and variances in mixed graphical association models
- Lauritzen
- 1992
(Show Context)
Citation Context ...ample [8], [7], [9], addresses this problem thoroughly and we utilize that theory here. In particular we incorporate Lauritzen's specialization of the BBN theory to the case of Gaussian distributions =-=[10]. A brief -=-account of the "evidence propagation" algorithm is given here, but we assume the reader is familiar with the basic ideas. It is well known that, if a probability is specified on a DAG, margi... |

123 |
An on-line algorithm for real-time accompaniment
- Dannenberg
(Show Context)
Citation Context ...ands of ensemble playing. And it would make music more accessible to the amateur player. Several other researchers have addressed this automatic accompaniment problem, most notably Dannenberg et. al. =-=[1]-=-,[2],[3] and Vercoe et. al. [4], [5] have had substantial success. In fact, Dannenberg 's research has even spawned a commercially available program. There are two significant distinctions between our... |

100 | Automatic Segmentation of Acoustic Musical Signals Using Hidden Markov Models
- Raphael
- 1999
(Show Context)
Citation Context ... "Listen" and "Play" form the heart of our system. The Listen process uses a hidden Markov model to track the evolution of the soloist's position within the score; this work is des=-=cribed in detail in [6]-=- and will not be discussed here. When the Listen process determines that a solo note has occurred it communicates this information to the Play process. Most often the detection of a note occurs shortl... |

56 |
The Synthetic performer in the context of live performance
- Vercoe
(Show Context)
Citation Context ...ld make music more accessible to the amateur player. Several other researchers have addressed this automatic accompaniment problem, most notably Dannenberg et. al. [1],[2],[3] and Vercoe et. al. [4], =-=[5]-=- have had substantial success. In fact, Dannenberg 's research has even spawned a commercially available program. There are two significant distinctions between our work and that of Dannenberg and Ver... |

39 |
Following an improvisation in real time
- Dannenberg, Mont-Reynaud
- 1987
(Show Context)
Citation Context ...re that has not been exploited elsewhere in applications to music, to our knowledge. Other examples of attempts at some form of computer-generated musical accompaniment are Dannenerg and Mont-Reynaud =-=[16]-=- which deals with accompanying an improvised solo line; Bryson [14], [15], which addresses chordal analysis; and Baird et. al. [19] which generates a real time accompaniment using MIDI input. Our syst... |

26 | Real-Time Computer Accompaniment of Keyboard Performances
- Bloch, Dannenberg
- 1985
(Show Context)
Citation Context ... of ensemble playing. And it would make music more accessible to the amateur player. Several other researchers have addressed this automatic accompaniment problem, most notably Dannenberg et. al. [1],=-=[2]-=-,[3] and Vercoe et. al. [4], [5] have had substantial success. In fact, Dannenberg 's research has even spawned a commercially available program. There are two significant distinctions between our wor... |

15 |
Artificial Intelligence and Music: Implementing an Interactive Computer Performer
- Baird, D, et al.
- 1993
(Show Context)
Citation Context ...puter-generated musical accompaniment are Dannenerg and Mont-Reynaud [16] which deals with accompanying an improvised solo line; Bryson [14], [15], which addresses chordal analysis; and Baird et. al. =-=[19]-=- which generates a real time accompaniment using MIDI input. Our system is composed of a chain of hardware components that conceptually form a loop. The actual input to the system comes from a microph... |

11 | The Reactive Accompanist: Adaptation and Behavior Decomposition in a Music System
- Bryson
- 1995
(Show Context)
Citation Context ... knowledge. Other examples of attempts at some form of computer-generated musical accompaniment are Dannenerg and Mont-Reynaud [16] which deals with accompanying an improvised solo line; Bryson [14], =-=[15]-=-, which addresses chordal analysis; and Baird et. al. [19] which generates a real time accompaniment using MIDI input. Our system is composed of a chain of hardware components that conceptually form a... |

8 | The subsumption strategy development of a music modeling system
- Bryson
- 1992
(Show Context)
Citation Context ...to our knowledge. Other examples of attempts at some form of computer-generated musical accompaniment are Dannenerg and Mont-Reynaud [16] which deals with accompanying an improvised solo line; Bryson =-=[14]-=-, [15], which addresses chordal analysis; and Baird et. al. [19] which generates a real time accompaniment using MIDI input. Our system is composed of a chain of hardware components that conceptually ... |

3 |
Mukaino "New Techniques for Enhanced Quality of Computer Accompaniment
- Dannenberg, H
- 1988
(Show Context)
Citation Context ...ensemble playing. And it would make music more accessible to the amateur player. Several other researchers have addressed this automatic accompaniment problem, most notably Dannenberg et. al. [1],[2],=-=[3]-=- and Vercoe et. al. [4], [5] have had substantial success. In fact, Dannenberg 's research has even spawned a commercially available program. There are two significant distinctions between our work an... |

3 |
Puckette "Synthetic Rehearsal: Training the Synthetic Performer
- Vercoe, M
- 1985
(Show Context)
Citation Context ...t would make music more accessible to the amateur player. Several other researchers have addressed this automatic accompaniment problem, most notably Dannenberg et. al. [1],[2],[3] and Vercoe et. al. =-=[4]-=-, [5] have had substantial success. In fact, Dannenberg 's research has even spawned a commercially available program. There are two significant distinctions between our work and that of Dannenberg an... |

3 |
A music scene analysis system with the MRF-based information integration scheme
- Kashino, Hagita
- 1996
(Show Context)
Citation Context ...enberg's does not. While we do not share any specific methodology or application, there is a strong philosophical connection between our work and that of Beran and Mazzola [13] and Kashino and Hagita =-=[18]-=-; in both cases, statistical methodology brings power and structure that has not been exploited elsewhere in applications to music, to our knowledge. Other examples of attempts at some form of compute... |

2 |
Analyzing Musical Structure and Performance- A statistical Approach
- Beran, Mazzola
- 1999
(Show Context)
Citation Context ...'s interpretation while Dannenberg's does not. While we do not share any specific methodology or application, there is a strong philosophical connection between our work and that of Beran and Mazzola =-=[13]-=- and Kashino and Hagita [18]; in both cases, statistical methodology brings power and structure that has not been exploited elsewhere in applications to music, to our knowledge. Other examples of atte... |

1 | ENA: Experimental Neural Accompanist - Camurri, Capocaccia, et al. - 1990 |