Search

Which spectral cues underlie the perceptual processing of smiles in speech? Here, the question was addressed using reverse-correlation in the case of the isolated vowel [a]. Listeners were presented with hundreds of pairs of utterances with randomly manipulated spectral characteristics and were asked to indicate, in each pair, which was the most smiling. The analyses revealed that they relied on robust spectral representations that specifically encoded vowel's formants. These findings demonstrate the causal role played by formants in the perception of smile. Overall, this paper suggests a general method to estimate the spectral bases of high-level (e.g., emotional/social/paralinguistic) speech representations.

The "bag-of-frames" (BOF) approach, which encodes audio signals as the long-term statistical distribution of short-term spectral features, is commonly regarded as an effective and sufficient way to represent environmental sound recordings (soundscapes). The present paper describes a conceptual replication of a use of the BOF approach in a seminal article using several other soundscape datasets, with results strongly questioning the adequacy of the BOF approach for the task. As demonstrated in this paper, the good accuracy originally reported with BOF likely resulted from a particularly permissive dataset with low within-class variability. Soundscape modeling, therefore, may not be the closed case it was once thought to be.

The paper describes an application of machine learning techniques to identify expiratory and inspiration phases from the audio recording of human baby cries. Crying episodes were recorded from 14 infants, spanning four vocalization contexts in their first 12 months of age; recordings from three individuals were annotated manually to identify expiratory and inspiratory sounds and used as training examples to segment automatically the recordings of the other 11 individuals. The proposed algorithm uses a hidden Markov model architecture, in which state likelihoods are estimated either with Gaussian mixture models or by converting the classification decisions of a support vector machine. The algorithm yields up to 95% classification precision (86% average), and its ability generalizes over different babies, different ages, and vocalization contexts. The technique offers an opportunity to quantify expiration duration, count the crying rate, and other time-related characteristics of baby crying for screening, diagnosis, and research purposes over large populations of infants.

This study uses an audio signal transformation, splicing, to create an experimental situation where human listeners judge the similarity of audio signals, which they cannot easily categorize. Splicing works by segmenting audio signals into 50-ms frames, then shuffling and concatenating these frames back in random order. Splicing a signal masks the identification of the categories that it normally elicits: For instance, human participants cannot easily identify the sound of cars in a spliced recording of a city street. This study compares human performance on both normal and spliced recordings of soundscapes and music. Splicing is found to degrade human similarity performance significantly less for soundscapes than for music: When two spliced soundscapes are judged similar to one another, the original recordings also tend to sound similar. This establishes that humans are capable of reconstructing consistent similarity relations between soundscapes without relying much on the identification of the natural categories associated with such signals, such as their constituent sound sources. This finding contradicts previous literature and points to new ways to conceptualize the different ways in which humans perceive soundscapes and music.

The "bag-of-frames" approach (BOF) to audio pattern recognition represents signals as the long-term statistical distribution of their local spectral features. This approach has proved nearly optimal for simulating the auditory perception of natural and human environments (or soundscapes), and is also the most predominent paradigm to extract high-level descriptions from music signals. However, recent studies show that, contrary to its application to soundscape signals, BOF only provides limited performance when applied to polyphonic music signals. This paper proposes to explicitly examine the difference between urban soundscapes and polyphonic music with respect to their modeling with the BOF approach. First, the application of the same measure of acoustic similarity on both soundscape and music data sets confirms that the BOF approach can model soundscapes to near-perfect precision, and exhibits none of the limitations observed in the music data set. Second, the modification of this measure by two custom homogeneity transforms reveals critical differences in the temporal and statistical structure of the typical frame distribution of each type of signal. Such differences may explain the uneven performance of BOF algorithms on soundscapes and music signals, and suggest that their human perception rely on cognitive processes of a different nature.