Hierarchical linear dynamical systems for unsupervised musical note recognition

Cinar, Goktug T.; Sequeira, Pedro M.; Prı́ncipe, José C.

doi:10.1016/j.jfranklin.2017.04.013

Cited by 3 publications

(6 citation statements)

References 42 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…We use the HLDS architecture from [1]. HLDS imposes a constraint that effectively regularizes the Kalman specifically for note segmentation.…”

Section: The Proposed Methodsmentioning

confidence: 99%

“…2) Relative stationarity is also controlled by setting the innovation variances of higher layers smaller compared to lower ones (Section III-B). The specific block structure of the joint transition matrix pioneered by [1] provides efficient online updates (using Kalman filtering) to the hierarchy of hidden states of the process.…”

Section: The Proposed Methodsmentioning

confidence: 99%

“…HLDS models time series using a set of hierarchical linear systems [1]. Without loss of generality and for ease of presentation, assume L = 2 where L denotes the number of hidden layers.…”

Section: B Hldsmentioning

confidence: 99%

“…These mappings are to be compared to where test observations get mapped to. We still utilize Kalman filtering during training and testing to calculate states {x t }, associated with each train or test observation, in order to extract {z t } as in [1]. Mappings {z t } can be exploited in many ways (next section).…”

Section: Training and Testingmentioning

confidence: 99%

See 3 more Smart Citations

Hierarchical Linear Dynamical System for Representing Notes from Recorded Audio

Kalantari¹,

Prı́ncipe²,

Sieving³

2022

Preprint

View full text Add to dashboard Cite

We seek to develop simultaneous segmentation and classification of notes from audio recordings in presence of outliers. The selected architecture for modeling time series is hierarchical linear dynamical system (HLDS). We propose a novel method for its parameter setting. HLDS can potentially be employed in two ways: 1) simultaneous segmentation and clustering for exploring data, i.e. finding unknown notes, 2) simultaneous segmentation and classification of audio recording for finding the notes of interest in the presence of outliers. We adapted HLDS for the second purpose since it is an easier task and still a challenging problem, e.g. in the field of bioacoustics. Each test clip has the same notes (but different instances) as of the training clip and also contain outlier notes. At test, it is automatically decided to which class of interest a note belongs to if any. Two applications of this work are to the fields of bioacoustics for detection of animal sounds in audio field recordings and also to musicology. Experiments have been conducted for segmentation and classification of both avian and musical notes from recorded audio.

show abstract

“…We use the HLDS architecture from [1]. HLDS imposes a constraint that effectively regularizes the Kalman specifically for note segmentation.…”

Section: The Proposed Methodsmentioning

confidence: 99%

Section: The Proposed Methodsmentioning

confidence: 99%

“…HLDS models time series using a set of hierarchical linear systems [1]. Without loss of generality and for ease of presentation, assume L = 2 where L denotes the number of hidden layers.…”

Section: B Hldsmentioning

confidence: 99%

Section: Training and Testingmentioning

confidence: 99%

See 2 more Smart Citations

Hierarchical Linear Dynamical System for Representing Notes from Recorded Audio

Kalantari¹,

Prı́ncipe²,

Sieving³

2022

Preprint

View full text Add to dashboard Cite

show abstract

“…HLDS is an architecture for linear modeling of time series data proposed by [9]. Later, we adapted HLDS for multiscale time series representation [8].…”

Section: Hldsmentioning

confidence: 99%

Uncertainty quantification for multiclass data description

Kalantari,

Principe,

Sieving

2021

Preprint

View full text Add to dashboard Cite

In this manuscript, we propose a multiclass data description model based on kernel Mahalanobis distance (MDD-KM) with self-adapting hyperparameter setting. MDD-KM provides uncertainty quantification and can be deployed to build classification systems for the realistic scenario where out-of-distribution (OOD) samples are present among the test data. Given a test signal, a quantity related to empirical kernel Mahalanobis distance between the signal and each of the training classes is computed. Since these quantities correspond to the same reproducing kernel Hilbert space, they are commensurable and hence can be readily treated as classification scores without further application of fusion techniques. To set kernel parameters, we exploit the fact that predictive variance according to a Gaussian process (GP) is empirical kernel Mahalanobis distance when a centralized kernel is used, and propose to use GP's negative likelihood function as the cost function. We conduct experiments on the real problem of avian note classification. We report a prototypical classification system based on a hierarchical linear dynamical system with MDD-KM as a component. Our classification system does not require sound event detection as a preprocessing step, and is able to find instances of training avian notes with varying length among OOD samples (corresponding to unknown notes of disinterest) in the test audio clip. Domain knowledge is leveraged to make crisp decisions from raw classification scores. We demonstrate the superior performance of MDD-KM over possibilistic K-nearest neighbor.

show abstract

Wavelet-packets Associated with Support Vector Machine Are Effective for Monophone Sorting in Music Signals

Scalvenzi

Guido

Marranghello

2019

Int. J. Semantic Computing

View full text Add to dashboard Cite

An abstract interpretation is usually required to analyze acoustic compositions. Nevertheless, there is much signal processing-related research focusing on music processing and similar topics. In that context, the semantic information contained in the melody involving major and minor chords, sharps and flats associated with semibreve, minim, crotchet, quaver, semiquaver and demisemiquaver notes can help in the study of musical sounds. Thus, multiresolution analysis based on discrete wavelet-packet transform (DWPT) associated with a support vector machine (SVM) is used in this paper to inspect and classify those signals, correlating them with a respective acoustic pattern. Results over hundreds of inputs provided almost full accuracy, reassuring the efficacy of the proposed approach for both off-line and real-time usage.

show abstract

Hierarchical linear dynamical systems for unsupervised musical note recognition

Cited by 3 publications

References 42 publications

Hierarchical Linear Dynamical System for Representing Notes from Recorded Audio

Hierarchical Linear Dynamical System for Representing Notes from Recorded Audio

Uncertainty quantification for multiclass data description

Wavelet-packets Associated with Support Vector Machine Are Effective for Monophone Sorting in Music Signals

Contact Info

Product

Resources

About