Improving Speech Understanding in Noisy Group Conversation: 86% of Participants Performed Better with Signia Integrated Xperience Versus Key Competitor

Learning Outcomes

After this course, learners will be able to:

• Describe the basic concept of Signia’s multi-stream architecture (RealTime Conversation Enhancement technology).
• Describe how the ability to understand speech in a group conversation in noise can be assessed in a study done in the laboratory.
• Explain the perceptual benefits offered by Signia IX in dynamic group conversations in noise.

Abstract

This article presents findings from an external validation study conducted by Hörzentrum Oldenburg, Germany. The study compared the Signia Integrated Xperience (IX) hearing aid with a competitor device featuring an AI co-processor. Results demonstrated significantly improved speech understanding with Signia IX in a test simulating a noisy group conversation. Notably, 86% of participants achieved better results with Signia IX compared to the competitor. These findings independently confirm previous studies, clearly demonstrating that technical signal-to-noise ratio advantages of Signia IX consistently translate into superior speech understanding for wearers.

Introduction

As every hearing care professional (HCP) who fits hearing aids knows, participating in group conversations in noise is one of the most challenging listening situations for people with hearing loss. Consequently, it is often the situation where hearing aid wearers require the most assistance from their hearing aids.

The challenges faced by individuals with hearing loss in group conversations are reinforced by numerous research findings. For example, in a survey of almost 15,000 people, in which most respondents had a self-reported hearing loss, “hearing friends and family in noise” was found to be the most desirable hearing aid attribute, with 88.3% of respondents rating this attribute to be very or extremely important (Manchaiah et al., 2021). This result corresponds well with MarkeTrak 22 data, which showed that hearing aid wearers reported their lowest level of listening satisfaction in noisy group conversations (Picou, 2022). Despite a general increase in hearing aid satisfaction over the last decade, driven largely by technology advances in rechargeability, direct streaming and smartphone apps, listening in multi-talker settings remains a challenge for many wearers.  

For most people, the ability to participate and contribute to conversations is highly important. Research consistently shows that difficulties in these types of listening situations negatively impact social participation and quality of life (e.g., Shukla et al., 2020; Heffernan et al., 2022).  These results underscore the critical need for effective solutions. Further, Nicoras et al. (2023) explored the elements contributing to successful conversations and identified "being able to listen easily" as the most crucial factor for both individuals with hearing loss and those with normal hearing. Interestingly, both groups considered this factor more vital in group conversations than in one-on-one interactions.

Traditional hearing aids often fall short of addressing the auditory challenges wearers face in a group conversation. While many premium hearing aids include directional microphones and noise reduction algorithms, their ability to adapt to the real-time complexities of group conversations remains limited. In dynamic social settings, wearers frequently report feeling overwhelmed or disconnected, as the technology fails to prioritize and dynamically adapt to the most relevant speech signals amidst the noise.

Recognizing these limitations, Signia introduced the Signia Integrated Xperience (IX) platform featuring RealTime Conversation Enhancement (RTCE) technology. This groundbreaking innovation, based on a multi-stream architecture, represents a significant advancement in hearing aid functionality, specifically designed to improve speech understanding in noisy group settings.

The perceptual benefits of activating RTCE have already been investigated and documented in several studies, conducted both in the lab and in the real world (e.g., Jensen et al., 2023; Folkeard et al., 2024; Korhonen & Slugocki, 2024; Slugocki et al., 2024a; Slugocki et al., 2024b). A relevant question is: How does the Signia IX multi-stream processing compare to the single-stream processing approaches used by competitors? The question has become even more relevant after the recent launches of competitor hearing aids with Artificial Intelligence (AI)-based noise reduction solutions, claimed to improve wearers’ speech-in-noise performance.

When developing appropriate test setups for benchmark comparisons, it’s essential to consider, as outlined above, that one of the most relevant and challenging scenarios for hearing aid wearers is a dynamic group conversation in background noise. This represents a particularly stringent benchmark for hearing aids, and one we believe should be the standard of comparison for evaluating hearing aids in realistic and demanding situations, rather than relying on static single-speaker lab settings.

To accurately replicate this type of scenario in a lab, certain key conditions must be met. Target speech should alternate from different directions to reflect the dynamic turn-taking of multiple talkers in a real-world conversation. Additionally, the background noise must be realistic, including a mix of both environmental sounds and distracting speech. Demonstrating benefits under these dynamic and realistic conditions provides greater confidence that the advantages of Signia IX will translate effectively to real-world use.

In a recent technical study, the signal-to-noise ratio (SNR) performance of Signia IX was compared to that of four key competitor hearing aids. The hearing aids were placed on a KEMAR in a simulated group conversation in background noise, following the above-mentioned test guidelines. The results demonstrated a clear advantage for Signia IX, which provided a 3.2 dB higher output SNR than the closest competitor. For more details on the study, see Jensen et al. (2024b).

While an SNR improvement measured at the hearing aid output is often linked to improved speech understanding, other factors may affect the real-world speech perception of the hearing aid wearer. For example, individual variations in hearing loss, auditory processing capabilities, and attention mean that a given change of the SNR may have a different impact on speech understanding for different listeners. Accordingly, to fully assess speech understanding in acoustically challenging settings, hearing aids must be assessed in studies with human participants.

To investigate whether the observed technical SNR benefit of Signia IX translates into a meaningful speech understanding benefit, different studies have been conducted to compare the speech understanding performance of Signia IX in a noisy group conversation scenario to that of the best-performing competitor from the previously cited study by Jensen et al. (2024b). Notably, this competitor was also the most recently launched hearing aid at the time (August, 2024), featuring an AI co-processor offering a Deep Neural Network (DNN)-based noise reduction system.

In a study performed by ORCA Labs in Chicago, Signia IX improved speech understanding in a simulated noisy group conversation, and participants reported that Signia IX increased the average time they were willing to spend in a very noisy group conversation by 85% (Korhonen et al., 2025). In another study performed in Signia’s Wonderful Sound Lab, a 1.4 dB improvement of the speech reception threshold, corresponding to a 22% improvement of speech understanding, was observed in a speech test simulating a group conversation in noise (Jensen et al., 2024a).

In this article, we report the results of an external study, which was undertaken by a third party (Hörzentrum Oldenburg, Germany), and which also compared Signia IX to the competitor with an AI co-processor. Before reviewing the study and its findings, we first provide a brief description of RTCE technology.

Real-Time Conversation Enhancement

This section offers a quick introduction to the RealTime Conversation Enhancement technology, which is implemented in Signia IX. More detailed information on the background and function of RTCE is provided by Jensen et al., 2023.

When the Signia IX wearer finds themselves in a group conversation, an overarching goal of RTCE is to improve the access to the speech of all conversation partners in noisy environments, allowing the wearer to focus on the talkers, while keeping the surrounding sound environment as comfortable as possible. RTCE works on top of the Augmented Focus split processing, which was introduced on the Signia Augmented Xperience (AX) platform (Branda, 2021). The split processing facilitates the three unique stages of RTCE, which are referred to as Analyze, Augment and Adapt. While these three stages are discussed separately to facilitate explanation, in reality these stages are interconnected, as illustrated simplistically in Figure 1, and they function continuously and simultaneously.

Figure 1. RealTime Conversation Enhancement consists of three interconnected stages – Analyze, Augment and Adapt – that function continuously and simultaneously.

In the first stage, Analyze, the complete conversation layout is determined. The analysis includes localization of the talkers involved in the conversation and assessment of the conversation dynamics. In the second stage, Augment, advanced binaural processing is used to create three independent focus streams, which are added to the existing focus stream from the split processing. This allows talkers in the conversation to be enhanced, while background noise is processed independently as a surrounding stream. In the third stage, Adapt, the streams are combined to rapidly adapt and keep up with changes in the conversation layout, e.g., when conversation partners take turns speaking, when they move or even when the wearer turns their head.

The ability to detect and enhance the different talkers in a group conversation and seamlessly adapt to any changes in the conversation layout in real time allows the wearer to focus on and contribute to the conversation. As opposed to most traditional hearing aids, RTCE technology does not need a long time for adaptive systems to settle before the wearer gets the performance boost they need in the given situation. The adaptation takes place instantly, allowing the wearer to always remain in the conversation and converse naturally.

As mentioned in the Introduction, the technical and perceptual benefits offered by RTCE have been demonstrated in numerous studies, including the latest presented below.

Methods

The study was conducted at Hörzentrum Oldenburg, Germany, and it was approved by the medical ethics committee of the University of Oldenburg.

Participants

An a priori statistical power analysis was performed using G*Power 3.1.9.7 to determine the appropriate number of participants. The analysis showed that 24 participants would allow detection of a 1 dB difference in the speech reception threshold (SRT) measured with the OLSA speech test (see below), with a significance level of 5% and a statistical power of 90%, and assuming a standard deviation of 1.4 dB on the individual SRT differences, based on data from previous use of the test.

To account for changes in the test setup and possible dropouts among the participants, a total of 30 potential participants were recruited for the study. All participants signed an informed consent form before entering the study.

One participant decided to drop out of the study without completing the testing, and one participant could not perform the test task reliably. Thus, a total of 28 participants completed the study and contributed to the data presented in this article.

The 28 participants included 16 females and 12 males. Their mean age was 74 years (range: 24-84 years), and they all had a mild-to-moderate-severe sensorineural hearing loss, with the mean audiogram shown in Figure 2. They were all experienced hearing aid wearers, wearing a wide variety of hearing aid brands in their daily life.

Figure 2. Mean audiogram for the left and right ear for the 28 participants. Error bars indicate ± one standard deviation.

Hearing Aids

All participants were fitted bilaterally with two pairs of receiver-in-canal (RIC) hearing aids: Signia Pure C&G T 7IX and a competitor’s newly launched premium hearing aid with an AI co-processor. As mentioned in the Introduction, this specific product was the best-performing competitor hearing aid in the technical benchmark study, in which Signia IX outperformed four different competitors’ hearing aids (Jensen et al., 2024b). This specific product was the same hearing aid used in a previous human performance comparison study (Jensen et al., 2024a). In accordance with the nomenclature used in those studies, the competitor hearing aid will be referred to as “Brand A1” in this article.

Both hearing aids were fitted individually according to the NAL-NL2 rationale (adult, non-tonal, experienced wearer, 100% acclimatization), and all fittings were done with closed couplings (power domes/sleeves). All processing channels were enabled, with frequency compression/lowering features deactivated. Frequency-specific fine-tuning of the gain was not allowed, but changes in the master gain were allowed if the prescribed gain setting could not be tolerated by the participant. However, this only happened in a few cases. The volume controls were disabled to keep the gain fixed during the following testing. During testing, both test hearing aids were set in a manual speech-in-noise program that assured all speech enhancement and noise reduction features, including DNN-based noise reduction, were always active.

Speech Test

All participants were subjected to a speech test that assessed their ability to understand conversational speech with the two different test hearing aids. The test setup simulated a group conversation in a noisy canteen and the test approach was based on the approach used in a previous study (Jensen et al., 2024a). However, in this study, a different room with different dimensions and acoustic characteristics was used, making the test conditions quite different than in the previous study.

The test was conducted in a meeting/conference room, and the loudspeaker setup used for the test is shown in Figure 3. A canteen background scenario was established by playing canteen noise from eight loudspeakers, spatially separated by 45° and positioned 2.62 m from the participant. The canteen noise was presented at a total level of 65 dBA (measured at the position of the participant, without the participant being present). To simulate a group conversation in the canteen scenario, target speech was presented alternating from two loudspeakers at 0° and 330° (both positioned 1.3 m from the participant). Additionally, two loudspeakers at 135° and 225° (both positioned 0.81 m from the participant) were used to simulate interfering speech from nearby talkers by presenting continuous speech produced by a female English talker and the ISTS signal (Holube et al., 2010). The presentation levels of the interfering speech signals were 70 dBA and 71 dBA, respectively. The total noise level of the canteen noise and the interfering speech was 74 dBA at the position of the participant (without the presence of the participant).

Figure 3. Loudspeaker setup used to simulate a group conversation involving three people. The participant was seated in the center of the setup, and target sentences were presented alternately from the two red loudspeakers, interfering speech noise was presented from the two gray loudspeakers, while ambient canteen noise was presented from the eight white loudspeakers. The distances to the loudspeakers are measured from the middle of the participant’s head (without the participant being present).

The test method used to assess the participants’ ability to understand speech in the conversation scenario was the standardized German Oldenburger Satztest (OLSA; Wagener et al., 1999). The target speech was the German OLSA sentences presented alternating from the two target loudspeakers to simulate the turn-taking in a conversation. During the test, the participant was allowed to move their head freely, as they would in a normal real-world conversation. The task of the participant was to repeat each sentence verbally. An experimenter scored the number of correctly repeated words, and depending on this number, the level of the subsequent sentence was changed adaptively to determine the signal-to-noise ratio (SNR) where 50% of the words could be repeated correctly, typically referred to as the speech reception threshold (SRT50).

Prior to data collection, a training round was completed to familiarize the participants with the OLSA test procedure. The training was done with both test hearing aids to introduce the sound of both test hearing aids to the participants prior to the actual testing. After completing the training, the participants were tested with both hearing aids, using 30 OLSA sentences for each condition. The test order of the hearing aids was counterbalanced across participants.

Results

The outcome of the test was individual SRT50 values for the 28 participants, obtained with Signia IX and the Brand A1 hearing aids, respectively. Figure 4 shows the mean SRT50 for both hearing aids. It should be noted that lower SRT50 values indicate better speech-in-noise performance.

Figure 4. Mean SRT50 for Signia IX and the Brand A1 competitor hearing aids, across the 28 participants. The error bars indicate one standard deviation. *** indicates p < .001.

The mean SRT50 across the 28 participants was -5.4 dB for Signia IX, while it was -3.9 dB for the competitor hearing aids. That is, Signia IX provided an SNR benefit of 1.5 dB in this conversation scenario. This benefit is statistically significant according to a two-sided paired t-test (t(27) = 5.17, p = .000019).

The SRT50 benefit translates directly to a benefit in speech understanding. The SRT50 difference can be converted to a difference in speech understanding when the slope of the performance-intensity (P-I) function (the psychometric function) is known, and when linearity of the P-I function within the SNR range of interest is assumed. By applying the slope value of 16 percent/dB for the OLSA test (for speech-shaped masking noise), which was published by Wagener & Brand (2005), the 1.5 dB SRT50 benefit of Signia IX corresponds to a speech understanding improvement of 24% (percentage points). This magnitude of speech understanding improvement is not only statistically significant, but also clinically significant.

The 28 individual SRT50 differences were calculated by subtracting the SRT50 for Signia IX from the SRT50 for Brand A1. In this way, a positive value of the difference indicates a speech understanding benefit of Signia IX, while a negative value indicates a benefit of Brand A1. The individual differences are plotted in the bar graph shown in Figure 5, where the participants are ordered according to the rank of the individual difference. In the plot, the red bars indicate the cases where the difference exceeded 0 dB, that is, where participants performed better with Signia IX.

Figure 5. Individual SRT50 differences. Positive values indicate a benefit of Signia IX (marked in red), while negative values indicate a benefit of Brand A1 (marked in gray). Participants have been numbered according to the rank of their individual difference.

As shown in Figure 5, 24 out of the 28 participants – 86% of the participants – performed better in the test with Signia IX than with Brand A1. The figure also shows that higher magnitudes of benefit were obtained with Signia IX, improving the SRT50 by as much as 5.6 dB and in most cases by more than 1 dB. In comparison, among the four participants performing better with Brand A1, the highest observed benefit was 1.2 dB whereas the benefit for the remaining three participants was below 1 dB.

Discussion

The results from this study show a clear speech understanding in noise benefit of Signia IX with RealTime Conversation Enhancement when compared to a competitor hearing aid with an AI co-processor-driven platform. As explained in the Results section, the observed SRT50 benefit of 1.5 dB with Signia IX corresponds to a 24 percentage-point increase in speech understanding.

The results of the study are consistent with the outcome of the technical comparison of the two hearing aids on the KEMAR (Jensen et al., 2024b), which showed Signia IX provided a higher output SNR in a simulated group conversation in noise compared to the Brand A1 hearing aids. While there may not be an exact one-to-one relationship between an output SNR improvement and the corresponding improvement in speech understanding (as measured by the SRT50), improving the SNR is an important prerequisite for improving speech understanding, and the effect is clearly demonstrated in this study.

Furthermore, the results are consistent with a previous study (Jensen et al., 2024a) that compared the speech understanding performance provided by Signia IX and Brand A1 in a comparable simulated conversation scenario. The previous study showed an SRT50 benefit of 1.4 dB, which is very close to the 1.5 dB observed in the present study. Thus, despite differences in test room acoustics and loudspeaker configuration, the present study basically replicates the findings from the previous study, suggesting a robust perceptual benefit of RTCE. This new finding strengthens the shared conclusion of the studies: Signia IX provides significantly better speech-in-noise performance than Brand A1 in the noisy group conversation scenario, which was investigated in different environments in a replicable manner.

When examining why Signia IX excels in noisy group conversations, three key features of its signal processing stand out. Firstly, the RTCE multi-stream processing approach used in Signia IX is fundamentally different from the conventional single-stream processing used by Brand A1 and other competitors, which process all incoming sounds in the same way.

Secondly, Signia IX's split processing allows for independent handling of the speech the wearer wants to focus on, separate from the surrounding sounds which are processed differently (e.g., with more noise reduction and compression). This creates a clear distinction between speech and background noise, unlike the single-stream approach used by competitors. As a result, the wearer can hear speech clearly while still being aware of their surroundings.

Thirdly, RTCE conducts a detailed analysis of the conversation scenario, directing multiple focus streams towards relevant conversation partners. This allows the wearer to easily access speech from more than one talker, even if the talkers move or the wearer turns their head. The ability to detect and track multiple conversation partners while reducing background noise gives Signia IX wearers a significant advantage in group conversations in noisy environments.

Summary

This study, conducted at Hörzentrum Oldenburg in Germany, evaluated the effectiveness of Signia IX and a premium competitor RIC hearing aid with an AI co-processor-driven platform in enhancing speech understanding during a simulated noisy group conversation. The assessment utilized a modified version of the standardized Oldenburger Satztest (OLSA), where target speech was presented from alternating directions amidst canteen noise and nearby interfering speech to mimic a realistic dynamic group conversation in noise.

The study included 28 participants with hearing loss. The findings revealed the mean speech reception threshold (SRT50) was 1.5 dB better (lower) with Signia IX than with the competitor’s hearing aids. This reduction in SRT50 is statistically significant and corresponds to a 24% increase in speech understanding. Notably, 86% of the participants performed better in the test with Signia IX compared to the competitor hearing aids.

This study confirms that the technical output SNR improvement of 3.1 dB provided by Signia IX compared to the competitor with an AI co-processor-driven platform – which was observed in a previous study – translates into a tangible speech understanding benefit for wearers in noisy group conversations. The results of this study highlight that the multi-stream processing offered by Signia IX's RealTime Conversation Enhancement significantly enhances speech comprehension in dynamic group conversation settings with background noise.

References

Branda, E. (2021). Split-processing: A new technology for a new generation of hearing aid. Audiology Practices, 13(4), 36–41.

Folkeard, P., Jensen, N. S., Kamkar Parsi, H., Bilert, S., & Scollie, S. (2024). Hearing at the mall: Multibeam processing technology improves hearing group conversations in a real-world food court environment. American Journal of Audiology, 33, 782–792. https://doi.org/10.1044/2024_AJA-24-00027

Heffernan, E., Withanachchi, C. M., & Ferguson, M. A. (2022). ‘The worse my hearing got, the less sociable I got’: A qualitative study of patient and professional views of the management of social isolation and hearing loss. Age and Ageing, 51(2), 1–10. https://doi.org/10.1093/ageing/afac019

Holube, I., Fredelake, S., Vlaming, M., & Kollmeier, B. (2010). Development and analysis of an international speech test signal (ISTS). International Journal of Audiology, 49(12), 891–903. https://doi.org/10.3109/14992027.2010.506889

Jensen, N. S., Mohnlein-Gilbert, K., Wilson, C., Berwick, N., Kamkar Parsi, H., Best, S., Samra, B., & Taylor, B. (2024a). Signia IX with pioneering multi-stream technology delivers 22% better speech understanding in noisy group conversation than a competitor with an AI co-processor-driven platform. Signia White Paper. Retrieved from www.signia-library.com

Jensen, N. S., Samra, B., Kamkar Parsi, H., Bilert, S., & Taylor, B. (2023). Power the conversation with Signia Integrated Xperience and RealTime Conversation Enhancement. Signia White Paper. Retrieved from www.signia-library.com

Jensen, N. S., Wilson, C., Kamkar Parsi, H., Samra, B., Hain, J., Best, S., & Taylor, B. (2024b). Signia IX delivers more than twice the speech enhancement benefit in a noisy group conversation than the closest competitors. Signia White Paper. Retrieved from www.signia-library.com

Korhonen, P., Kuk, F., Slugocki, C., & Peeters, H. (2025). Conversations in noise: Multi-stream architecture vs. deep neural network approach to hearing aids. Hearing Review, 32(1), 18–21.

Korhonen, P., & Slugocki, C. (2024). Augmenting split processing with a multi-stream architecture algorithm. Hearing Review, 31(5), 20–23.

Manchaiah, V., Picou, E. M., Bailey, A., & Rodrigo, H. (2021). Consumer ratings of the most desirable hearing aid attributes. Journal of the American Academy of Audiology, 32(8), 537–546. https://doi.org/10.1055/s-0041-1732442

Nicoras, R., Gotowiec, S., Hadley, L. V., Smeds, K., & Naylor, G. (2023). Conversation success in one-to-one and group conversation: A group concept mapping study of adults with normal and impaired hearing. International Journal of Audiology, 62(9), 868–876. https://doi.org/10.1080/14992027.2022.2095538

Shukla, A., Harper, M., Pedersen, E., Goman, A., Suen, J. J., et al. (2020). Hearing loss, loneliness, and social isolation: A systematic review. Otolaryngology and Head and Neck Surgery, 162(5), 622–633. https://doi.org/10.1177/0194599820910377

Slugocki, C., Kuk, F., & Korhonen, P. (2024a). Using alpha-band power to evaluate hearing aid directionality based on multistream architecture. American Journal of Audiology, 33(4), 1164–1175. https://doi.org/10.1044/2024_AJA-24-00117

Slugocki, C., Kuk, F., & Korhonen, P. (2024b). Using the mismatch negativity to evaluate hearing aid directional enhancement based on multistream architecture. Ear and Hearing, Early Online, 1–11. https://doi.org/10.1097/AUD.0000000000001619

Wagener, K., Kühnel, V., & Kollmeier, B. (1999). Entwicklung und Evaluation eines Satztests für die deutsche Sprache I: Design des Oldenburger Satztests [Development and evaluation of a sentence test for the German I: Design of the Oldenburg sentence test]. Zeitschrift für Audiologie (Audiological Acoustics), 38, 4–15.

Wagener, K. C., & Brand, T. (2005). Sentence intelligibility in noise for listeners with normal hearing and hearing impairment: Influence of measurement procedure and masking parameters. International Journal of Audiology, 44(3), 144–156. https://doi.org/10.1080/14992020500057517