Abstract not found

Few types of signal streams are as ubiquitous as music. Here we consider the problem of extracting essential ingredients of music signals, such as well-defined global temporal structure in the form of nested periodicities (or meter). Can we construct an adaptive signal processing device that learns by example how to generate new instances of a given musical style? Because recurrent neural networks can in principle learn the temporal structure of a signal, they are good candidates for such a task. Unfortunately, music composed by standard recurrent neural networks (RNNs) often lacks global coherence. The reason for this failure seems to be that RNNs cannot keep track of temporally distant events that indicate global music structure. Long Short-Term Memory (LSTM) has succeeded in similar domains where other RNNs have failed, such as timing & counting and learning of context sensitive languages. In the current study we show that LSTM is also a good mechanism for learning to compose music. We present experimental results showing that LSTM successfully learns a form of blues music and is able to compose novel (and we believe pleasing) melodies in that style. Remarkably, once the network has found the relevant structure it does not drift from it: LSTM is able to play the blues with good timing and proper structure as long as one is willing to listen.

The thesis proposes an algorithm for the recognition of musical meter from acoustic signals of music. Musical meter is a part of rhythm that is constantly present in music, as it spans the musical time base. The proposed model is capable of finding metrical levels, including the beat and the tatum, in real time from a musical audio signal. The model comprises four main components: an onset detector, a tatum estimator, a phenomenal accent model, and a beat estimator. The onset detector finds distinct sound onsets from an acoustic signal, using multiband signal processing. After this, the tatum, which is the lowest metrical level, is computed from onset times. Phenomenal accents are computed from a set of 16 acoustic signal features using Bayesian pattern recognition. The tatum and the accents then yield the beat. The proposed model operates causally and is able to respond to tempo changes. The design of the model aims at generality in regard to musical genres, and thus the model is trained and tested using 330 music excerpts from multiple genres. The model performance varies according to the rhythmic difficulty of the input signal. Most pop/rock music poses no problems for the algorithm, while classical music and expressive jazz pieces are intractable. The model produces more errors than Eric Scheirer's beat tracker, but at the same time it follows more metrical levels than Scheirer's model. The results of this thesis are directly applicable in music production and post-processing. The access to musical time enables new levels of productivity and automation in both music software and hardware. Meter-synchronized comparison, mixing, and editing of pieces of music is possible. Robust meter recognition is a vital component of music information retrieval applications.

The Normalized Positive (NPOS) model is a rule-based model that predicts downbeat location and pattern complexity in rhythmical patterns. Though derived from several existing models, the NPOS model is particularly effective at making correct predictions while at the same time having low complexity. In this paper, the details of the model are explored and a comparison is made to existing models. Several datasets are used to examine the complexity predictions of the model. Special attention is paid to the model’s ability to account for the effects of musical experience on beat induction. 1

In general music composed by recurrent neural networks (RNNs) suffers from a lack of global structure. Though networks can learn note-by-note transition probabilities and even reproduce phrases, attempts at learning an entire musical form and using that knowledge to guide composition have been unsuccessful. The reason for this failure seems to be that RNNs cannot keep track of temporally distant events that indicate global music structure. Long Short-Term Memory (LSTM) has succeeded in similar domains where other RNNs have failed, such as timing & counting and CSL learning. In the current study we show that LSTM is also a good mechanism for learning to compose music. We compare this approach to previous attempts, with particular focus on issues of data representation. We present experimental results showing that LSTM successfully learns a form of blues music and is able to compose novel (and we believe pleasing) melodies in that style. Remarkably, once the network has found the relevant structure it does not drift from it: LSTM is able to play the blues with good timing and proper structure as long as one is willing to listen.

this report. This thesis will describe attempts to build a real-time interactive system for computer music which allows a human instrumentalist to establish the rhythmic frame of performance. The computer must search for evidence of metrical hierarchy or timeline from the audio signal, tracking tactus (beat induction) whilst allowing for expressive timing and motor noise, and all this without excessive zeal but within some reasonably smooth trajectory