Stimuli and speech tests

Each of the processing conditions were evaluated using two speech perception experiments. The first part measured the perception of speech in quiet (SIQ), without background noise. This part used the IEEE speech corpus⁴⁹, spoken by a British male talker, which consists of 72 lists with 10 sentences with 5 keywords each. Speech perception was assessed for SIQ as the percentage of keywords identified correctly for 2 lists presented per condition (20 sentences). Lists were chosen randomly without replacement.

The second part measured the perception of speech in noise (SIN). In contrast to the previous part, the SIN part used the BKB speech corpus⁵⁰ in combination with the 20-talker babble noise from Auditec (St. Louis). The BKB sentences, used clinically in the UK, are typically easier than the IEEE sentences and were used to assess performance with a different speech corpus in addition to the IEEE sentences. The BKB lists consist of 15 sentences with 3 keywords each, spoken by a British male talker. The SIN test procedure⁵¹ used an adaptive one-up/one-down procedure to measure the speech reception threshold (SRT) at which 50% of the sentences were understood correctly. Starting with an initial signal-to-noise ratio (SNR) of 4 dB SNR, the procedure used a randomly selected sentence and a stepsize of 2 dB per trial to alter the noise level until the participant repeated the three keywords of the first sentence correctly. It then started with the adaptive one-up/one-down procedure to determine the SRT for that list as the average of the last ten SNRs presented. Two lists of 15 sentences were presented for each condition and the final SRT score was calculated as the average of the two SRTs.

During the experiment, participants took off their own clinical CI speech processor and used an Advanced Bionics research speech processor connected with an auxiliary cable to an external sound card (Roland Quad-Capture) and laptop computer (Dell XPS 15). The fitting procedure for the research speech processor parameters and the programming of the experimental MAPs onto the research speech processor was performed using BEPS + software from Advanced Bionics. The research software for the speech tests was programmed in MATLAB (The Mathworks).