Speech_DPOAEs

This repository contains the complete dataset from "Otoacoustic emissions evoked by the time-varying harmonic structure of speech" paper, and an example code covering the analysis method and leading to figure 3E. It contains: (1) The pure-tone and speech otoacousitc emissions (OAEs) of the 24 subjects (Recordings_1.zip; 13 subjects and out of the ear control, Recordings_2.zip; 11 remaining subjects). (2) The speech derived waveforms used to generate as well as to measure the speech-OAEs (Waveforms_Male.zip; male voice, Waveforms_Female.zip; female voice). (3) The raw speech stimuli from which the waveforms were derived and to which subjects payed attention to (Speech.zip). (4) The basic Matlab analysis code leading to the speech-OAE function (Code.zip) which includes a function for reading the recording files, the analysis pipeline and the computation of the complex cross-correlation. (5) Readme.txt file describes how to run the example code. Stimuli The speech stimuli are continuous natural speech (audiobooks) recorded by us. A male and female speaker read the same story. The audiobook is divided into 2 minutes sections (section_1 to section_28) and so are the derived harmonic waveforms, both for male and female voices. Recordings The individual recordings contain (1) the pure-tone DPOAEs (i.e. using primary frequencies of 1 kHz and 1.2 kHz, for 30 seconds) and (2) the speech-OAEs. The speech-OAEs recordings are divided into 14 sections (not to be confused with the stimuli and waveforms sections). As the paper describes, the speech-OAEs were recorded from the contralateral ear by simulating the right ear with two waveforms wm(t) and wn(t) that tracked the mth and nth harmonics of the fundamental frequency of the speech we wanted to record the speech-OAEs from, and which was simultaneously played in the left ear. The speech-OAEs were measured while the subjects listed to speech both in quiet and in noise. To this end, we employed different 2 minutes segments. Some speech segments consisted only of the male or of the female voice, while others had both voices mixed together. There was a total of 14 segments (section_1 to section_14 in individual recordings). Key information is detailed in results_details.txt files. For each subject and segment, results_details.txt, describes the recording condition. In particular, it says which gender (male or female) and speech/waveform section (1 to 28) it was used. We started by measuring speech-OAEs in quiet. To this end we employed one speech stimulus that consisted of the male voice, as well as another stimulus that contained the female voice. These correspond to section_1 and section_2, respectively. Therefore, results_details.txt files in these sections say “DPOAE gender: Male, DPOAE section: 1, Attended: one speaker”, and “DPOAE gender: Female, DPOAE section: 2, Attended: one speaker”, respectively, meaning that male, and female voices in quiet were tracked, using segments 1 and 2 from the audiobooks. The potential influence of selective attention to speech on the speech-OAES was then assessed. We therefore employed 12 segments (sections_2 to section_14 in recordings) with competing speech, that is, segments that contained both the male and the female voice. During each segment the subject was asked to attend either the female or the male speaker. Speech-OAEs related to either the male or the female speaker were measured from the contralateral ear. The order of the attentional focus, as well as the order in which speech-OAEs to the male and the female voice were measured, was determined randomly per subject. Speech_permultation.txt file describes this information, for each participant. For instance, for subject 21 this file says: “Attended: Female_ Female_ Female* Female* Male_ Male_ Male_ Male* Female* Male* Male* Female_”. It indicates that the subject was instructed to pay attention to that order of speakers along the 12 segments. The starts (*) denote that, the speech-OAEs were measured in those segments, while the underscore (_) denotes that the competing story was tracked. The attended vs ignored segments were paired. They were all played and recorded twice, once as attended and one as ignored such that we obtained three recordings of speech-OAEs to the male voice, both when that voice was attended and when it was ignored; and analogously with the female condition. To get this information, one has to look at the results_details.txt files from the individual segments. As an example, subject 21, section_5, reads “DPOAE gender: Female, DPOAE section: 5, Attended: Yes”. It means that we recorded the speech-OAEs of the female voice reading audiobook section 5, while the subject was paying attention to it. This competing speakers segment would be compared with section_9 which reads “DPOAE gender: Female, DPOAE section: 5, Attended: No”, as the same audiobook section was used but this time the attention was focused in the competing speaker. Refer to readme.txt for further clarification.