Spectral analysis of infant cries and adult speech

• Published in: International journal of speech technology Vol. 19; no. 4; pp. 841 - 856
• Main Authors: Chittora, Anshu, Patil, Hemant A.
• Format: Journal Article
• Language: English
• Published: New York Springer US 01.12.2016; Springer Nature B.V
• Subjects: Adults; Artificial Intelligence; Classification; Classifiers; Crying; Energy distribution; Engineering; Fourier transforms; Infants; Signal classification; Signal,Image and Speech Processing; Social Sciences; Spectrograms; Spectroscopy; Spectrum analysis; Speech; Support vector machines; STFT; Spectrographic analysis; Infant cry; SVM classifier
• ISSN: 1381-2416; 1572-8110
• DOI: 10.1007/s10772-016-9375-z

In this paper, spectrographic analysis of the infant cries is reported. For the spectrographic analysis of the infant cries ten different cry modes are used to analyze differences in different pathological cries. A comparison of spectrograms of the adult speech signal and infant cry signals is given. Based on differences in the distribution of energy in the spectrograms, energy-based features are calculated from the short-time Fourier transform (STFT) of the adult speech and infant cry signals. The classification performance of these features is obtained using support vector machine (SVM) classifier and it is observed that the energy distribution in 0–1 kHz range is promising feature in the classification of adult speech and infant cries and the classification accuracy achieved with this feature is 98.22 %. On the contrary, it was observed that it is very difficult to classify adult speech and infant cries using the energy distribution in 1–3 kHz.