Anjali Jayakumar

A. Jayakumar, T. Bhattacharjee, S. Vengalil, Y. Belur, N. Atchayaram, and P. K. Ghosh, “Low complexity model with single dimensional feature for speech based classification of amyotrophic lateral sclerosis patients and healthy individuals,” in International Conference on Signal Processing and Communications (SPCOM). IEEE, 2024, pp. 1–5.

Plots for mel spectrogram of Amyotrophic Lateral Sclerosis (ALS) and Healthy speech samples for Spontaneous (SPON) and Diadochokinetic rate (DIDK) tasks

Feature distribution for ALS and Healthy Class for Spontaneous (SPON) and Diadochokinetic rate (DIDK) tasks

The features considered are the standard deviation (SD) and Autocorrelation with lag 1 and 2 (AC(1) and AC(2)) of the 1st and 2nd coefficients of MFCC and its delta and double delta, over a dataset comprising of speech samples from 72 ALS and 55 healthy subjects.

SPON Task

DIDK Task

ALS VS Healthy Classification Metrices for Spontaneous (SPON) and Diadochokinetic rate (DIDK) tasks

DNN models with varying numbers of dense layers (DNN-2L, DNN-1L, DNN-0L) are compared against a baseline CNN-BiLSTM model using different input features. Training and validation utilize speech samples from 72 ALS and 55 healthy subjects via 5-fold cross-validation. Mean value of True Positive Rate (TPR), True Negative Rate (TNR), False Positive Rate (FPR), and False Negative Rate (FNR) over the 5 folds are plotted. The error bars and the shaded regions represent the standard deviation over the 5 folds.

SPON Task

DIDK Task

Page updated

Google Sites

Report abuse