Singing Transcription from Polyphonic Music Using Melody Contour Filtering

Automatic singing transcription and analysis from polyphonic music records are essential in a number of indexing techniques for computational auditory scenes. To obtain a note-level sequence in this work, we divide the singing transcription task into two subtasks: melody extraction and note transcription. We construct a salience function in terms of harmonic and rhythmic similarity and a measurement of spectral balance. Central to our proposed method is the measurement of melody contours, which are calculated using edge searching based on their continuity properties. We calculate the mean contour salience by separating melody analysis from the adjacent breakpoint connective strength matrix, and we select the final melody contour to determine MIDI notes. This unique method, combining audio signals with image edge analysis, provides a more interpretable analysis platform for continuous singing signals. Experimental analysis using Music Information Retrieval Evaluation Exchange (MIREX) datasets shows that our technique achieves promising results both for audio melody extraction and polyphonic singing transcription.

Musical Intervals in Infants’ Spontaneous Crying over the First 4 Months of Life

<b><i>Introduction:</i></b> Perception and memorizing of melody and rhythm start about the third trimester of gestation. Infants have astonishing musical predispositions, and melody contour is most salient for them. <b><i>Objective:</i></b> To longitudinally analyse melody contour of spontaneous crying of healthy infants and to identify melodic intervals. The aim was 3-fold: (1) to answer the question whether spontaneous crying of healthy infants regularly exhibits melodic intervals across the observation period, (2) to investigate whether interval events become more complex with age and (3) to analyse interval size distribution. <b><i>Methods:</i></b> Weekly cry recordings of 12 healthy infants (6 females) over the first 4 months of life were analysed (6,130 cry utterances) using frequency spectrograms and pitch analyses (PRAAT). A preselection of utterances containing a well-identifiable, noise-free and undisturbed melodic contour was applied to identify and measure melodic intervals in the final subset of 3,114 utterances. Age-dependent frequency of occurrence of melodic intervals was statistically analysed using generalized estimating equations. <b><i>Results:</i></b> 85.3% of all preselected melody contours (<i>n</i> = 3,114) either contained single rising or falling melodic intervals or complex events as combinations of both. In total 6,814 melodic intervals were measured. A significant increase in interval occurrence was found characterized by a non-linear age effect (3 developmental phases). Complex events were found to significantly increase linearly with age. In both calculations, no sex effect was found. Interval size distribution showed a maximum of the minor second as the prevailing musical interval in infants’ crying over the first 4 months of life. <b><i>Conclusion:</i></b> Melodic intervals seem to be a regular phenomenon of spontaneous crying of healthy infants. They are suggested to be a further candidate for developing an early risk marker of vocal control in infants. Subsequent studies are needed to compare healthy infants and infants at risk for respiratory-laryngeal dysfunction to investigate the diagnostic value of the occurrence of melodic intervals and their age-depending complexification.

Classification-Based Singing Melody Extraction Using Deep Convolutional Neural Networks

Singing melody extraction is the task that identifies the melody pitch contour of singing voice from polyphonic music. Most of the traditional melody extraction algorithms are based on calculating salient pitch candidates or separating the melody source from the mixture. Recently, classification-based approach based on deep learning has drawn much attentions. In this paper, we present a classification-based singing melody extraction model using deep convolutional neural networks. The proposed model consists of a singing pitch extractor (SPE) and a singing voice activity detector (SVAD). The SPE is trained to predict a high-resolution pitch label of singing voice from a short segment of spectrogram. This allows the model to predict highly continuous curves. The melody contour is smoothed further by post-processing the output of the melody extractor. The SVAD is trained to determine if a long segment of mel-spectrogram contains a singing voice. This often produces voice false alarm errors around the boundary of singing segments. We reduced them by exploiting the output of the SPE. Finally, we evaluate the proposed melody extraction model on several public datasets. The results show that the proposed model is comparable to state-of-the-art algorithms.