Task T1:
Development of a programming environment for
content extraction and sound manipulation.
Final score: we studied and developed musical description
components and components of sound transformation.
Subtask T1.1: Study of musical content description
components.
Final score: we precisely determined the required
functionalities of
the musical content analysis components, in the context of
the previous
research group developments. In particular, we focused on
the
description of similarity, expressivity and musical
structure.
Subtask T1.2: Study of physical transformation of sound
components.
Final score: we accurately determined the required
functionalities for
the physical transformation of sound components.
Subtask T1.3: Development of components for the
description of musical
content.
Final score: we developed a library of sound analysis
components and
description of high-level musical content apropiate to the
problems
tackled in this project.
Subtask T1.4: Development of physical transformation of
sound components.
Final score: we developed a library of content-based
transformation
components for musical recordings.
Subtask T1.5: Development of a programming language for
the manipulation
of sound analysis and transformation components.
Final score: we developed a programming language for
the manipulation of sound analysis and transformation
components.
Task T2:
Pattern recognition techniques for sequences analysis.
Final score:
1. We developed and evaluated symbolic
analysis
techniques for the selection and extraction of useful
descriptors to
recover musical information.
2. We developed and evaluated pattern recognition techniques tested them
using the project
dataset.
Subtask T2.1: Corpora acquisition.
Final score: we achieved, described and organized data built
corpora for the experiments. It was necessary to have corpora
both for
the specific domain of the project and for other reference
domains.
Subtask T2.2 Design and evaluation of data
representations.
Final score: we defined and studied different models to
represent
symbolic music information.
Subtask T2.3: Adaptive metrics.
Final score: we modified the costs of different edit
distances to
improve their performance. This can be done by accomplishing these
objectives:
1) Extension of the conditional stochastic edit
distance to
the transducers with memory.
2) Extension of the conditional
stochastic edit distance to measure differences between hierarchic
structures.
3) To get a procedure to find the most probable representatives
of a set of patterns using the stochastic edit distance.
Subtask T2.4: Analysis of dynamic and cooperative
classification techniques.
Final score: we studied and evaluated different techniques
to combine classifiers, in order to combine different classification
paradigms and to combine partial decisions. To evaluate the feasibility of
the partial decisions which allow an online categorization of the data
(while they are entering into the system), to facilitate that
applications to work in real time.
Subtask T2.5: Efficiency improvement to make
decisions.
Final score: To get scalability efficiency (for large
databases), due to the high computational cost of the metrics (for example,
the edit distance between trees) and because it is necessary that
some applications of the project will work in real time.
Task T3:
Demonstration.
Final score: Design of a prototype to show the Project
research results.
Subtask T3.1: Extension of the database of musical
recordings.
Final score: we generated a corpus of recorded musical
performances meeting the special requirements of this project.
Subtask T3.2: Design of a musical transcription
prototype.
Final score: we developed a prototype able to
extract the notes from a digital audio recording, by means of signal
processing and spectral pattern recognition techniques.
Subtask T3.3: Prototype design of music content
descriptors.
Final score: we developed prototypical musical
content descriptors processing recorded performances. This goal achieved by means of
spectral analysis and spectral models.
Subtask T3.4: Design of a physical sound
transformation prototype.
Final score: we designed a prototype allowing musical
content transformation of a monophonic recording
Subtask T3.5: Prototype implementation of a
programming language for the
manipulation of musical-content description
components.
Final score: we implemented a prototype of a concurrent
programming language for the manipulation of the sound analysis
components and the sound transformation operations provided by the subtasks
T1.3 and T1.4.
Subtask T3.6: Incorporation of new developments of
similarity and
recognition measures to the melody processing
prototype.
Final score: to incorporate the techniques developed
in task T2 to
the prototype of processing and categorization of digital
scores. These
techniques will involve: advances in the metrics
developed in subtask
T2.3, the algorithms for classifier combination and
cooperative and
dynamic decisions developed in subtask T2.4, and the
methodologies able
to decide in an efficient manner from subtask T2.5.
Subtask T3.7: Development of a final
prototype.
Final score: we developed the final prototype of a
helper tool for the handling of musical recordings in the context of
sound post-production.
Task T4:
Monitoring and evaluation.
Final score: monitoring, control and evaluation of the
different stages of the project. Adaptation of the working plan to
the possible eventualities. Coordination of diffusion and exploitation
activities.
