Prediction of Negative Symptoms of Schizophrenia from Emotion Related Low-Level Speech Signals

Abstract

Negative symptoms of schizophrenia are often associated with the blunting of emotional affect which creates a serious impediment in the daily functioning of the patients. Affective prosody is almost always adversely impacted in such cases, and is known to exhibit itself through the low-level acoustic signals of prosody. To automate and simplify the process of assessment of severity of emotion related symptoms of schizophrenia, we utilized these low-level acoustic signals to predict the expert subjective ratings assigned by a trained psychologist during an interview with the patient. Specifically, we extract acoustic features related to emotion using the openSMILE toolkit from the audio recordings of the interviews. We analysed the interviews of 78 paid participants (52 patients and 26 healthy controls) in this study. The subjective ratings could be accurately predicted from the objective openSMILE acoustic signals with an accuracy of 61-85% using machine-learning algorithms with leave-oneout cross-validation technique. Furthermore, these objective measures can be reliably utilized to distinguish between the patient and healthy groups, as supervised learning methods can classify the two groups with 79-86% accuracy.

Speech databases for mental disorders: A systematic review

“…The cognitive models applied the Addenbrooke’s Cognitive Examination-Revised, the picture naming task (International Picture Naming Project, IPNP), the National Adult Reading Test (IPNP), and so on, accompanied by a one-time MRI scan and an in-depth neuropsychological assessment. In addition, Chakraborty and colleagues22 investigated into patients with schizophrenia who were unable to identify emotion from voice due to their incapability in understanding low-level acoustic features, such as pitch, intensity, frequency, and so forth. The Negative Symptom Assessment-16 scale23 rating method evaluated the behavior of those 78 participants (52 patients and 26 controls).…”

Sch-net: A Deep Learning Architecture for Automatic Schizophrenia Setection

“…1) Pitch-related features: Pitch is the fundamental frequency of vocal cord vibration for voiced initial consonants and some unvoiced initial consonants [35]. Pitchrelated features are common used in analyzing flat affect of negative symptoms [24,25,26,27,28,29,30,3]. Studies [28,27,26,24] demonstrated that schizophrenia speech characterized by less variability in vocal pitch than normal speech.…”

“…Spectrum-related features are commonly used for schizophrenia speech analysis, reflecting the energy distribution and the vocal tract characteristics during speech production. The typical spectrum-related features, such as formants, auditory-based spectral features, and spectral envelope features, had been proved to be an effective feature for schizophrenia speech detection [28,33,29,34]. 1) Formants: Formant is the descriptor that reflects the resonance frequency of the vocal tract.…”

“…LPC is calculated to estimate the vocal tract model, which reflects the resonance during speech production. Studies [29,34] used MFCCs and LPCs to analyze the characteristics of schizophrenia speech. Results in [34] showed that the MFCC scores of schizophrenia speech are significantly lower than that of controls, and the LPC scores of schizophrenia speech are significantly higher than that of controls.…”

“…Generally, most existing methods [24,25,26,27,28,29,30,3,31,32,33,34] analyze schizophrenia speech using feature engineering and classifiers/correlation analysis. Schizophrenia speech detection experiments are conducted by extracting fluency features, intensity-related features, spectrum-related features, and so on.…”