Nineteen adult CI users aged between 15 and 69 years (average 42 years) participated in the study. We included CI users with at least 6-month unilateral implant experience and either a Cochlear Nucleus Freedom or a CP810 Sound Processor fitted with a frequency allocation table reaching higher than 5 kHz. Each test subject had a minimal speech perception score of 50% for German monosyllabic words at 60 dB sound pressure level (SPL), 3 months after implantation. Table 1 shows patient characteristics of the included subjects. The study protocol was fully approved by the local institutional review board. All patients gave written informed consent.

We compared Internet telephony (Skype, local area network connection) against the conventional telephone (PSTN, landline connection) using two devices, a personal computer (PC; Latitude E6510, Dell, Round Rock, TX, USA) and a cordless DECT telephone device (Philips VoIP855, Royal Philips Electronics, Amsterdam, The Netherlands) which has dual transmission functions (Skype app and PSTN). In addition, we compared four accessories (PC speakers Z320; Logitech headphones, Behringer HPS 500; 3.5 mm jack audio cable, cochlear Ltd; and induction loop) and five different Internet transmission speeds (controlled with a connection emulator, Perfect Soft Research, Version 1.3.2 Brisbane, Australia). We used Internet connections with random data PL in 5% steps ranging from 0% PL (perfect), 5% PL (mild), 10% PL (medium), 15% (severe) to 20% PL (very severe). Figure 1 shows the test setup with the two transmission modes (PSTN versus Skype) connected with two devices (PC, Figure 1 E and cordless DECT telephone, Figure 1 D).

The frequency response was measured for all devices using an audio analyzer (UPV, Rhode & Schwarz, Munich, Germany) and a head and torso simulator (KEMAR Manikin Type 45BA, Brüel & Kjaer, Naerum, Denmark). For acoustic measurements, the Manikin’s ear simulator (Type 4158) was used to simulate the situation of a telephone held on the ear. The ear simulator is composed of a silicon external ear (or pinna) coupled to an ear canal terminated by a half-inch condenser microphone and pre-amplifier. A sweep of 50 logarithmically spread, pure sinus tones was generated and the output was filtered with a 1% bandwidth filter locked on the stimulus frequency. The root mean square amplitude of the output was calculated to create a frequency-domain graph. Objective voice quality was tested using the audio analyzer’s built-in algorithm for Perceptual Evaluation of Speech Quality (PESQ) based on the guidelines of the ITU (International Telecommunication Union) in accordance with the ITU-T P.862 protocol. Test subjects rated the subjective perceived voice quality from 1-5 using the mean opinion score (MOS; Table 2, according to specifications ITU-T Rec. P.862.1 and P.862.2) [26].

The standardized German “HSM” sentence test [27] was used for open set monosyllable speech perception testing in noise (60 dB SPL) at a constant signal sound level of 70 dB SPL (free field at 1-m distance). The HSM test consists of 30 lists with 20 short sentences containing 106 monosyllable words. The order of the sentence was changed at random to avoid learning effects. The subject had to repeat the presented sentences, and received 1 point for each correctly reproduced word. The percent of speech perception from a total of 106 words was calculated for each condition (one list). Test subjects were tested monaurally in a sound treated room (Type 402A, Industrial Acoustics Company, Niederkrüchten, Germany), with an average reverberation time of 0.10 s (125-10000 Hz). The contralateral ear canal was closed by an earplug (USA EARlink 3C, EAR Corporation, Indianapolis) if there was a residual hearing. Other hearing aids or a second CI had to be turned off. Patients kept the everyday settings of their speech processor. The cordless DECT telephone (Figure 1 D) was coupled to the implant either with an audio cable (Personal Audio Cable, Cochlear Limited, Sydney, Australia) branched to the 3.5-mm jack socket of the handset or by holding it directly to the ear. We chose this experimental setting with speech signal presented in noise not only to simulate an everyday listening situation but also to avoid ceiling effects.

Robust nonparametric analyses were performed to assess the potentially non-normally distributed speech perception scores from this small study population. A two-tailed Wilcoxon matched pairs signed-rank test was used to compare Skype with PSTN. For the ideal condition with no Internet data PL (condition 0% PL), a one-tailed test was applied because of the expected superiority of Skype under this condition [21,23]. A P value<.05 was considered significant after applying a Bonferroni correction for multiple testing.

The differences of the transmitted frequencies of PSTN versus Skype (LAN connection) are shown in Figure 2. There was a better frequency response of the audio signal derived from the headphone jack compared with the telephone handset (Figure 2). A broader frequency range (50–5000 Hz) was transmitted via Skype than PSTN (200-3000 Hz) independently of whether the built-in loudspeaker or the headphone jack was used.

Speech perception with an optimal Skype connection (0% PL; median 91.6%, n=18, range 48.1-99.1%; interquartile range, IQR, 15.6) which was significantly better (P<.001) than the telephone speech perception (PSTN median 42.5%, n=18, range 11.3- 85.8%, IQR 15.6); however, there was statistically no advantage in speech understanding using Skype at lower Internet quality connections (PL 15% and 20%; Figure 3). Quality measurements of the transmitted sound (PESQ measurement) showed that all Skype quality connections (PL 0- 20%) offered a significantly better voice quality (P<.001) compared to PSTN (Figure 3). Skype voice quality was maintained even with a medium quality Internet connection (PL 5% and 10%; Figure 3)

Free-field speech perception with Skype (PC version) using an optimal Internet connection (0% PL) was 91.5 % (median, n=18, range 48.1-99.1 %, IQR 15.6). Speech perception with a connected induction loop was 79.3% (median, n=15, range 53.8-100%, IQR 17.92), with headphones 83.9 % (median, n=18, range 14.2-100 %, IQR 9.215), and with coupled CI cables 88.2 % (median, n=18, range 47.1-100 %, IQR 20.27). There was no significant difference (P>.05) between the tested accessories (Figure 4).

Figure 5 shows the speech perception performance with a DECT telephone coupled to the CI either with a cable or the handset. Median speech perception using a CI audio cable was 7.5% for PSTN (n=17, range 0-40.6%, IQR 21.7) compared with speech understanding of 65.1% with the Skype app (median, n=17, range 47.17-95.3%, IQR 30.2) installed on the DECT telephone. Speech understanding with Skype was significantly superior if the DECT telephone was directly coupled to the CI by cable (P<.001). There was, however, no significant advantage seen for Skype if the handset was held directly near the microphone above the pinna (Figure 5).

The voice quality measurements (PESQ) on the headphone jack socket of a DECT phone (cable connection) showed significantly higher scores for the installed Skype app (median 3.08, n=19, range 2.43- 3.71, IQR 0.285) compared with the PSTN connection (median 0.73, n=19, range 0.33-1.73, IQR 0.94, P<.001) on the same device (Figure 5). The built-in telephone loudspeakers of the handset, however, did not show any significant voice quality differences between PSTN and the Skype app (Figure 5).

The subjective perceived voice quality, which describes the level of effort required for understanding, was best with an optimal Skype connection (MOS 4.1 [SD 0.9] at PL 0%; Table 3; Figure 3). The audio cable was rated as the best accessory requiring the least effort for understanding speech (MOS 4.2, SD 0.7), while the neck loop was found to offer the most unpleasant sound (MOS 3.5, SD 1.2). Figures 3 and 5 show a comparison between psychoacoustic (panels A), technical (panels B), and subjective (panels C) performance.

Speech perception by CI patients using Skype with active loudspeakers was superior to the conventional telephone under perfect or medium network voice transmissions. There was no advantage of Internet telephony for transmissions with severe or very severe data PL (>10% PL). Accessories such as a neck loop with wireless telecoil, an audio cable directly connected to the CI or headphones, did not further improve speech perception; however, the CI users subjectively perceived the voice quality (MOS) as superior, when using direct cable input. A dedicated Skype DECT telephone did not offer the full advantage of the superior voice quality provided by the Skype app because of loudspeaker quality limitations.

A proof of concept was given by a previous experimental study showing a superior voice quality of Internet telephony resulting in better speech understanding, provided that the Internet connection quality was ideal [21]. This phenomenon was also shown in normal hearing subjects and was not related to any speech coding strategy of a CI [21].

Sound quality might be affected by low bit-rate coding, data PL, background noise, silence suppression, or by network filtering leading to sound delay, jitter, noise, and speech level changes. These parameters might not equally influence speech perception performance by CI users. The PESQ score, however, uses objective parameters to model psychoacoustic and cognitive perception of speech. Such a model was calibrated to predict MOS scores and to improve correlation between PESQ and MOS scores [26].

Live Skype calls (desktop version) transmitted via a deteriorated Internet connection (5-10% PL) still generated a better voice quality on the receiver side compared with a PSTN connection, but not for transmission modes with severe PL (15% and 20%). These findings are in line with speech perception test results performed under simulated laboratory conditions [20].

The main reason for the superiority of VoIP is likely due to technical reasons. The frequency coverage with VoIP is higher than that of conventional telephony (0.1-8 kHz vs 0.3-3.4 kHz; Figure 2) resulting in better audio quality (see PESQ measurements). Since Skype transmits higher frequencies, it is likely to convey more of the speech-relevant signal content such as consonants. In addition, telecommunication companies digitalize the analogue voice signal using low bit-rate coding (8 kHz sampling rate, G.711 codec, ITU recommendations, G-series) and maximum bit rates of 33.6 kbits/s compared with Skype, which uses a sampling rate of 16 kHz and variable bit rates up to 40 kbits/s [28].

No improvement in speech perception, however, was measured for the CI users when PL was >10% (Figure 3). This was unexpected after analyzing objective voice quality measurements (PESQ), which proved a better audio quality of Skype, regardless of the Internet connection quality (Figure 3). All tests were performed at a constant background noise, making these tests more sensitive to speech signal deterioration, which might have an impact on speech perception performance.

Accessories may improve telephone listening experience in background noise [29]. We measured no significant speech discrimination differences whether accessories were used or not. Possible ceiling effects might have occurred, since median speech perception performance was around 90% regardless of the accessory used. Test conditions with a lower signal-to-noise ratio (SNR < 10 dB) might have yielded different performance results since assisting listening accessories protect from unwanted noise by either shielding physically (around-the-ear/circumaural headphones) or by routing the signal directly to the device (cable, induction, FM system). Loudspeakers, however, do not offer an improved SNR, which is important for speech discrimination in CI users.

We tested only monaurally to simulate an everyday telephone use, and this might have affected the general performance with accessories. Studies show significant advantages in speech perception with binaural hearing (particularly in environmental noise) [30] or even in bimodal hearing [31]. Assistive listening devices enable users to transmit sound on both ears simultaneously.

Our data suggest that the choice of accessories should be made according to personal preference. Low quality built-in loudspeakers of the Skype DECT device resulted in lower voice quality and consequently in a lower speech perception performance. The Skype DECT device offered a better speech perception if the headphone jack was used. Most subjects also reported the best subjective listening experience with the audio cable connecting the headphone jack with the CI, although a cable connection limits the range of use. This might change with new wireless streaming possibilities [29,32,33], which have not been tested yet with VoIP applications.

This study analyzed telephone speech perception using real-time settings and measurements, which give a better estimate than the previous laboratory tests.

The patients kept the everyday settings of their speech processor, however, another way to improve the telephone listening experience is the fine structure preservation [6] or the application of a special telephone fitting mode [34] by reducing the current level for electrodes stimulating outside the transmitted frequency range.

Although all individuals were blinded regarding the tested condition, performance bias might still be possible, since individuals were aware of the used accessories or might have perceived the presented voice quality. The small sample size might lead to an under- or overestimation of study results. In addition, this is a self-controlled study without any other control group.

We tested one single VoIP software and selected accessories. These results are therefore not generalizable for all available VoIP programs or other Internet transmissions modes.

Further research is mandatory to test voice quality and speech perception performance through wireless and mobile Internet connections. New technologies such as wireless audio streaming [35] may further improve listening experience and performance. New generations of Bluetooth technology with low battery consumption and direct connection to the implant might replace any assisting telephone accessory in the future while preserving voice quality.

Internet telephony improves speech perception performance even under real and adverse network conditions. CI users who are not able to have a meaningful telephone conversation could improve their telephone listening experience by using Skype or any other broadband Internet telephony service. This might have a direct impact on social integration, general health, life expectancy [16], and cognitive function in the elderly [18,36,37]. Restrictions in interactions and activities because of hearing loss might result in reduction of the overall health status and thus increase morbidity and mortality [16].

Skype and other VoIP software are freely available and can be used with any computer, microphone and speaker system. Additional assisting listening devices and telephone accessories might help for binaural hearing or hearing in noise.

Broadband VoIP software such as Skype can significantly improve telecommunication experience for CI users even with low quality Internet connections. Listening accessories such as headphones, audio cables, or an induction loop were equivalent in terms of speech perception performance. Microphone and speaker quality of Skype telephone DECT devices do not fully exploit benefits of Skype apps which provide an enhanced broadband audio and voice quality.