Infected Phonemes: How a Cold Impairs Speech on a Phonetic Level

Abstract

The realization of language through vocal sounds involves a complex interplay between the lungs, the vocal cords, and a series of resonant chambers (e.g. mouth and nasal cavities). Due to their connection to the outside world, these body parts are popular spots for viruses and bacteria to enter the human organism. Affected people may suffer from an upper respiratory tract infection (URTIC) and consequently their voice often sounds breathy, raspy or sniffly. In this paper, we investigate the audible effects of a cold on a phonetic level. Results on a German corpus show that the articulation of consonants is more impaired than that of vowels. Surprisingly, nasal sounds do not follow this trend in our experiments. We finally try to predict a speaker's health condition by fusing decisions we derive from single phonemes. The presented work is part of the INTER-

Significance of voiced and unvoiced speech segments for the detection of common cold

“…Table 3 shows the performance comparison of the proposed framework with the state-of-the-art methods using various phoneme segmentation on the URTIC database. Wagner et al [18] analyze the effect of common cold speech on a phonetic level. Phonemes are grouped into vowel, glide, consonant, liquid, stops, fricative, and nasal sounds, and the classification performance is evaluated using MFCC, invariant-integration features (IIF), and constrained maximum likelihood linear regression (CMLLR) feature.…”

“…But UAR achieved using MFCC features extracted from VLR is low compared to the state-of-the-art methods. Wagner et al [18] examined the effects of common cold on phonetic level. They concluded that phoneme-level cold and healthy speech classification was not worthwhile.…”

“…used phoneme state posteriorgram (PSP) features for classifying cold and healthy speech. Wagner Flowchart for the segmentation of the voiced and unvoiced region et al[18] investigated the effect of common cold at phonetic level. Author derived a phonetic transcription from an automatic speech recognizer (ASR), and based on phonetic transcription, the author trained a classification model for each phonetic class using different low-level features such as MFCC, invariant-integration features (IIF), and constrained maximum likelihood linear regression (CMLLR).…”

Stress Annotations from Older Adults - Exploring the Foundations for Mobile ML-Based Health Assistance

“…In general, machine learning deals with the problem of designing algorithms that can automatically learn to detect patterns in data and allow a computer to make predictions based on what it has learned. In recent years machine learning methods have been applied successfully to solve various health-related recognition tasks [10,13,32] including stress detection [1,3,25]. Despite its continuously growing success the method is still subject to some caveats.…”