Predicting Speech Intelligibility Using a Gammachirp Envelope Distortion Index Based on the Signal-to-Distortion Ratio

Abstract

A new intelligibility prediction measure, called "Gammachirp Envelope Distortion Index (GEDI)" is proposed for the evaluation of speech enhancement algorithms. This model calculates the signal-to-distortion ratio (SDR) in envelope responses SDR env derived from the gammachirp filterbank outputs of clean and enhanced speech, and is an extension of the speech based envelope power spectrum model (sEPSM) to improve prediction and usability. An evaluation was performed by comparing human subjective results and model predictions for the speech intelligibility of noise-reduced sounds processed by spectral subtraction and a recent Wiener filtering technique. The proposed GEDI predicted the subjective results of the Wiener filtering better than those predicted by the original sEPSM and well-known conventional measures, i.e., STOI, CSII, and HASPI.

The gammachirp auditory filter and its application to speech perception

“…The dcGC-FB was also used in models for speech intelligibility prediction [9][10][11][12][13]. A new model referred to as GEDI (the gammachirp envelope distortion index) [10,11] predicted the intelligibility of speech sounds processed with non-linear enhancement algorithms better than other recent indexes like STOI, CSII, and HASPI [12,13].…”

“…We begin by defining the relationship between the ratio, f rat , and the total stimulus level at the output of the pGC, P gcp , as shown in Eq. (10). The slope of the ratio, f ð1Þ…”

“…This average level is used to determine the frequency ratio for the HP-AF, f rat , in Eq. (10) and, subsequently, the gain of the ''inverse'' HP-AF. The gain vector for the filterbank channels is used as the ''inverse'' excitation pattern.…”

“…The GC architecture enables us to construct a hearing impairment simulator which allows normal hearing listeners to experience the difficulties of hearing impaired listeners [5,6]. The GC has also been used to model speaker size perception [7,8] and speech intelligibility [9][10][11][12][13].…”

Multi-resolution Gammachirp Envelope Distortion Index for Intelligibility Prediction of Noisy Speech

“…The STOI is intended to assess the intelligibility of speech processed via an ideal time-frequency segregation (ITFS). It has been reported, however, that the STOI was not successful at predicting the intelligibility of speech sounds enhanced by via Wiener filtering [3] and a recent DNN-based enhancement algorithm [4].…”