2.1. Ear-EEG

Ear-EEG technology stands out by recording EEG signals from within the ear canal, representing a significant departure from traditional scalp EEG methodologies. Conventional scalp EEG requires an extensive setup process involving substantial time, bulky equipment, and limited suitability for daily wear. In contrast, ear-EEG offers an innovative, robust, discreet, and noninvasive solution for monitoring brain activity in real-world, non-laboratory settings (Mikkelsen et al., Reference Mikkelsen, Kappel, Mandic and Kidmose2015). This approach has gained increasing attention for its ability to capture EEG signals directly from the ear canal. However, the anatomical variability of the human ear poses significant challenges to the use of standardized ear-EEG electrodes, often resulting in inconsistent signal performance. Therefore, electrodes must either be flexible enough to conform to individual ear canal shapes or be manufactured in customized geometries. While wet electrodes remain the gold standard for EEG signal quality, their reliance on conductive gel makes them impractical for daily wearable use. Dry electrodes are more convenient but often suffer from signal degradation due to high skin impedance. Recent developments in semi-dry electrodes, such as hydrogel-coated designs, present a promising compromise, although further research is required to ensure their long-term stability and comfort (Ge et al., Reference Ge, Guo, Gong, Han, Feng, Ji, Sun, Gao, Bian and Xu2023). Figure 2 presents representative examples of current ear-EEG electrode designs, highlighting key variations in form factor, integration, and personalization.

Figure 2. Various types of ear EEG electrodes have been developed in the field. (a) Generalized electrode design based on a common ear-hub (Kidmose et al., Reference Kidmose, Looney, Jochumsen and Mandic2013). (b) Earhub-style electrode produced by molding (Sintotskiy and Hinrichs, Reference Sintotskiy and Hinrichs2020). (c) Earhub-style electrode with an integrated auricle holder (Looney et al., Reference Looney, Kidmose and Mandic2014). (d) Single-piece design combining the ear electrode and circuit board (Our ear-EEG) (L Yu et al., Reference Yu, Xu, Contreras and Kavehei2024). (e) 3D model of the integrated electrode and circuit board, featuring adjustable parameters including $ {D}_{\mathrm{o}} $ (outer diameter), $ {D}_{\mathrm{i}} $ (inner diameter), and $ L $ (length), as well as customizable electrode shape and curvature. This design enables on-demand manufacturing tailored to individual anatomies across different age groups and genders. The rigid structure ensures a durable and reliable product suitable for everyday life, for example, if the device drops on the ground or is pressed a little bit while in the pocket, and so forth.

To address these issues, generalizable ear-EEG electrodes have been developed to fit everyone. Such electrodes are typically constructed from elastic materials, enabling them to compress and expand to conform to the unique shape of an individual’s ear canal. This adaptability ensures a secure fit, enhancing signal quality and user comfort during prolonged use. This approach is favored due to its straightforward manufacturing process, which simplifies production while offering a “one-size-fits-all” solution. For instance, one design utilizes a foam substrate to allow the ear-EEG electrode to adapt to the shape of the ear canal, improving fit and signal stability (Goverdovsky et al., Reference Goverdovsky, Rosenberg, Nakamura, Looney, Sharp, Papavassiliou, Morrell and Mandic2017). This design features two electrodes and two microphones for applying stimulation. The choice of material effectively mitigates motion artifacts arising from both minor and substantial mechanical deformations of the ear canal walls (Goverdovsky et al., Reference Goverdovsky, Rosenberg, Nakamura, Looney, Sharp, Papavassiliou, Morrell and Mandic2017). This electrode design also incorporates the use of two microphones to capture acoustic signals traversing through dense head tissues and uses these signals. Another design utilized conductive silver fabric, foam earplugs, heat-shrinkable tubing, and copper wire to construct a generalized electrode (Juez et al., Reference Juez, Henao, Segura, Gómez, Le Van Quyen and Valderrama2021). To address the issue of long manufacturing time and short lifespan for personalized ear-EEG electrodes, Liang et al. developed a method using impressions to manufacture a generalized ear-EEG based on the average measurements of the ear at three critical dimensions: the aperture, isthmus, and length (Liang et al., Reference Liang, Wang, Li, Wang, Liu and Liu2023).

Researchers have increasingly employed three-dimensional (3D) printing technology to optimize EEG signal acquisition in the ear-EEG domain. By leveraging the capabilities of additive manufacturing, complex shapes can be created, enabling a precise fit within the ear canal for improved electrode contact and signal quality. Notable advancements include the work of Kaveh et al., who developed personalized wireless ear-EEG electrodes using 3D-printed molds combined with a silicone base (Kaveh et al., Reference Kaveh, Doong, Zhou, Schwendeman, Gopalan, Burghardt, Arias, Maharbiz and Muller2020). This approach allows for the fabrication of both customized and generalized electrodes. Similarly, Yu et al. introduced a 3D-printed ear-EEG device that integrates both the electrode and circuit board into a single printed structure (L Yu et al., Reference Yu, Xu, Contreras and Kavehei2024). This design has the potential to accommodate various ear shapes, enhancing usability across different individuals. Despite these innovations, the manufacturing process often requires additional steps, such as casting and chemical treatments, to apply conductive materials, making production more complex. To address these challenges, an ear-EEG electrode with an extendable spiral structure was developed, using electrothermal actuation to conform to the unique geometry of an individual’s ear canal (Z Wang et al., Reference Wang, Shi, Zhang, Zheng, Li, Jiao, Cheng, Wang, Zhang, Chen, Chen, Wang, Xie, Wang, Ma, Gao and Feng2023). This design, which explores applications in visual and auditory BCIs, features a dual-layer structure composed of shape memory polymers (SMPs). The outer electrothermal actuation layer (EAL) facilitates expansion, while the EEG detection layer (EEGDL) consists of Au wires and insulating polyimide for signal acquisition. Upon heat activation, the electrode spirals outward, adapting seamlessly to the unique curvature of the auditory meatus for a customized fit. Further contributions to 3D-printed ear-EEG technology include Jeong and Jeong, who created ear canal impressions with moldable plastic beads and applied silver paste for conductivity, and Mikkelsen et al., who developed soft silicone earplugs with silver buttons and copper wire connections for automatic sleep staging (Mikkelsen et al., Reference Mikkelsen, Villadsen, Otto and Kidmose2017; Jeong and Jeong, Reference Jeong and Jeong2020). Additionally, Tabar et al. employed 3D printing to fabricate customized ear-EEG electrodes, enhancing both user comfort and signal accuracy (Tabar et al., Reference Tabar, Mikkelsen, Rank, Hemmsen, Otto and Kidmose2021).

An alternative approach for creating personalized ear-EEG electrodes involves integrating standard electrodes into custom-moulded soft earpieces. One implementation employed a dry ear-EEG system featuring $ {\mathrm{IrO}}_2 $ -coated titanium electrodes embedded in elastic earmold silicone (Kappel et al., Reference Kappel, Rank, Toft, Andersen and Kidmose2018). Another high-density configuration used 15 electrodes per ear to achieve signal quality comparable to, or even exceeding, that of scalp EEG in certain evaluations, such as auditory steady-state response (ASSR) and steady-state visual evoked potential (SSVEP), also utilizing $ {\mathrm{IrO}}_2 $ as the electrode material (Kappel and Kidmose, Reference Kappel and Kidmose2017). In a separate design, a 3D-printed ear canal model was used to install six silver electrodes for objective threshold estimation tasks (Bech Christensen et al., Reference Bech Christensen, Hietkamp, Harte, Lunner and Kidmose2018).

Various materials have been explored to enhance ear-EEG performance. One notable example involves a porous-structured platinum electrode specifically designed for ear-EEG applications, which achieved an impedance below 5 $ \mathrm{k}\Omega $ at 20 Hz under dry conditions (Eickenscheidt et al., Reference Eickenscheidt, Schäfer, Baslan, Schwarz and Stieglitz2020).

Additional technologies, such as around-Ear-EEG systems, have been developed to investigate optimal electrode placement. Bleichner et al. created an around-Ear-EEG system using several layers of biocompatible polyimide, with conductive components consisting of gold-plated ends, pure copper traces, and conductive Ag/AgCl-based polymer thick film ink (Bleichner et al., Reference Bleichner, Mirkovic and Debener2016).

After gathering the data from the electrode, it is necessary to preprocess the signal and convert the signal from the analog domain into the digital domain, and send it to a host device for further applications. An analog front end is often necessary to convert the analog signal into digital form to enhance the ear-EEG devices’ portability. However, integrating the circuit board naturally and unobtrusively presents another challenge. Methods to reduce the device’s visibility are required. To address visibility concerns, one approach involved placing the analog front-end (AFE) module at the back of a hat, thereby keeping the system unobtrusive while maintaining signal fidelity (Bleichner et al., Reference Bleichner, Lundbeck, Selisky, Minow, Jäger, Emkes, Debener and De Vos2015).

There is an emerging need to integrate the analog front end directly onto the printed electrode to further augment the functionality and discretion of ear-EEG electrodes. This approach minimizes the electrode’s size and visibility while optimizing signal reception. Integrating the analog front end with the ear-EEG electrode makes it possible to achieve a less obtrusive and more efficient EEG recording system. This integration streamlines the manufacturing process and enhances the performance and user experience of ear-EEG technology.

Different sensors are also integrated into the ear-EEG piece to gather additional information for medical and research use. An in-ear integrated array of multimodal electrophysiological and electrochemical sensors has been developed to simultaneously sense EEG, Electrooculogram (EOG), and lactate concentration (Xu et al., Reference Xu, De la Paz, Paul, Mahato, Sempionatto, Tostado, Lee, Hota, Lin and Uppal2023). This demonstrates the feasibility of integrating multiple sensors into a single device, enabling the collection of diverse biosignals. Such a multi-sensor approach enhances the device’s capacity to provide comprehensive physiological data, facilitating advanced applications, such as personalized health monitoring, real-time diagnostics, and sophisticated data analysis, for medical and research purposes.

Seamlessly integrating the complete analog front end and the microcontroller (MCU) into a single earpiece to facilitate the wireless transmission of digital signals to a designated device remains a substantial challenge in this field. This challenge is further accentuated by the necessity of minimizing the distance between the electrode and the analog front end and the size of the auditory meatus (Mozaffari et al., Reference Mozaffari, Nash and Tucker2021). One suggested approach entails the strategic distribution of electrical components around the external ear rather than direct insertion into the ear canal (Kaveh et al., Reference Kaveh, Doong, Zhou, Schwendeman, Gopalan, Burghardt, Arias, Maharbiz and Muller2020). However, even with this approach, the visibility of these components remains noticeable, potentially causing social discomfort. In practical scenarios, this situation might require additional accessories, such as a hat, to effectively conceal the components.

It is important to acknowledge that while addressing this challenge entails inherent complexities, its resolution is attainable. Utilizing application-specific integrated circuits (ASICs) has emerged as a promising solution; however, it is pertinent to recognize that producing a limited quantity of ASICs can involve significant costs. As time progresses, the continued advancement of manufacturing and electronics technologies contributes to the ongoing miniaturization of electronic components. This enduring trend plays a pivotal role in facilitating the seamless integration of all necessary components within the confined space of the earpiece.

Ensuring that users can still capture environmental sounds is crucial for most electrodes. A hollow structure is often applied to ensure the patient can hear surrounding sounds effectively.

Table 2 compares various ear-EEG electrodes based on impedance, electrode material, signal type, and connectivity. The findings highlight that while dry electrodes improve convenience, wet electrodes still offer the best signal quality.

Table 2. Comparison between different ear-EEG electrodes

^a SSVEP, steady-state visual evoked potential.

^b ASSR, auditory steady-state response.

^c VEP, visual evoked potential.

^d BLE, bluetooth.

^e BCC, body channel communication.

2.1.2. Ear EEG recordings

The performance of ear-EEG has been evaluated in numerous studies, demonstrating that while its signal quality is generally lower than that of conventional scalp EEG, it remains sufficient for detecting reliable neural responses. This trade-off in signal fidelity is often considered acceptable given the substantial benefits of ear-EEG in terms of unobtrusiveness, comfort, and suitability for long-term, real-world monitoring.

Yu et al. validated the feasibility of ear-EEG for auditory stimulus detection using the ASSR paradigm (Yu et al., Reference Yu, Xu, Contreras and Kavehei2024). As shown in Figure 3, frequency-specific neural responses were clearly detected using ear-EEG with wet electrodes, demonstrating a strong correspondence to responses captured via scalp EEG. In both modalities, green lines indicate the expected response frequencies, while red lines mark the time points of stimulus frequency changes.

Figure 3. ASSR test using (a) scalp EEG and (b) ear-EEG. Red lines indicate the timing of frequency changes, while green lines mark the expected auditory response frequencies (Yu et al., Reference Yu, Xu, Contreras and Kavehei2024).

To further evaluate the capabilities of ear-EEG, we analyzed data from the Ear-EEG Sleep Monitoring 2019 (EESM19) dataset (Mikkelsen et al., Reference Mikkelsen, Kidmose and Rezaei Tabar2019). The recordings were collected using personalized dry IrO $ {}_x $ electrodes on 20 participants (7 males and 13 females) with an average age of 25.9 years (range: 23–36) (Bjarke Mikkelsen et al., Reference Bjarke Mikkelsen, Rezai Tabar, Rævsbæk Birch, Lind Kappel, Bech Christensen, Dalskov Mosgaard, Otto, Christian Hemmsen, Lind Rank and Kidmose2025). A TMSi Mobita amplifier was used for signal acquisition. For the ASSR test, they were exposed to a 40 Hz amplitude modulated at 1 kHz auditory stimulus (Bjarke Mikkelsen et al., Reference Bjarke Mikkelsen, Rezai Tabar, Rævsbæk Birch, Lind Kappel, Bech Christensen, Dalskov Mosgaard, Otto, Christian Hemmsen, Lind Rank and Kidmose2025). A corresponding neural response at 40 Hz was expected, representing the ASSR. Figure 4 shows the results across 20 subjects, comparing signal-to-noise ratio (SNR) measurements between scalp and ear-EEG. The SNR was defined as the power at 40 Hz relative to the average power within the 35–45 Hz band, excluding 40 Hz. Results demonstrate that ear-EEG consistently captured the 40 Hz ASSR, with performance comparable to that of scalp EEG.

Figure 4. Comparison of ASSR test results between scalp EEG and ear-EEG across 20 subjects from the EESM19 dataset, where the signal-to-noise ratio (SNR) is defined as the power at 40 Hz relative to the average power in the 35–45 Hz band, excluding 40 Hz (Mikkelsen et al., Reference Mikkelsen, Kidmose and Rezaei Tabar2019).

These data collectively support the viability of ear-EEG as a practical and reliable alternative to traditional scalp EEG, particularly for auditory stimulus monitoring in wearable applications.

In-ear EEG systems were generally well tolerated, with most participants reporting comfort and ease of use, even during sleep recordings. Minor issues such as earpiece displacement or pressure around the tragus were occasionally reported, and intermittent Bluetooth interruptions occurred in some cases (Moumane et al., Reference Moumane, Pazuelo, Nassar, Juez, Valderrama and Le Van Quyen2024). In contrast, scalp EEG often requires gel application and can cause discomfort over extended recordings, suggesting that in-ear EEG may offer advantages for long-term or ambulatory monitoring (Moumane et al., Reference Moumane, Pazuelo, Nassar, Juez, Valderrama and Le Van Quyen2024). Dry in-ear electrodes exhibited initially high impedance (~900 k $ \Omega $ ) that decreased and stabilized over ~3 h (~290 k $ \Omega $ at 50 Hz), comparable to other state-of-the-art dry electrodes. While scalp electrodes with gel maintain low impedance throughout, in-ear EEG achieves sufficient contact over time to ensure signal quality, highlighting its potential for prolonged, unobtrusive recordings. In-ear EEG successfully captured key rhythms, such as alpha, theta, spindles, and slow waves, with waveforms temporally aligned with scalp EEG. However, amplitudes were generally lower, about two times smaller for alpha waves, and fluctuations were slightly higher during naps due to movement or intermittent contact loss. Artifact prevalence was minimal during controlled alpha tests but increased during sleep, particularly with earplug displacement (Moumane et al., Reference Moumane, Pazuelo, Nassar, Juez, Valderrama and Le Van Quyen2024). Scalp EEG generally shows higher amplitude and more stable RMS values, but requires more preparation and is less comfortable. In-ear EEG reliably detected alpha peaks during eyes-closed conditions, although power was roughly half that of scalp recordings. SNRs were slightly lower, but correlations with scalp signals were high (93% of subjects), especially over temporal regions. Across sleep stages, relative spectral power in-ear closely matched scalp EEG, with small differences in alpha, delta, and beta1 bands during Wake. Correlations strengthened during deep sleep (N3), indicating coherent slow-wave activity in both modalities (Moumane et al., Reference Moumane, Pazuelo, Nassar, Juez, Valderrama and Le Van Quyen2024).

Ear-EEG demonstrates good temporal and spectral fidelity compared to scalp EEG, despite amplitude reductions and occasional susceptibility to artifacts. It is particularly promising for applications requiring long-term or ambulatory monitoring, where comfort and ease of use are critical (Moumane et al., Reference Moumane, Pazuelo, Nassar, Juez, Valderrama and Le Van Quyen2024). Scalp EEG remains superior in terms of absolute signal amplitude and stability, but in-ear EEG offers a practical trade-off between usability and recording quality (Moumane et al., Reference Moumane, Pazuelo, Nassar, Juez, Valderrama and Le Van Quyen2024).

2.2. Other applications

While ear-EEG remains the focus of most neurotechnology efforts, hearables have also been explored for other biosignals. Ear-ECG captures cardiac activity and may support heart rate variability analysis or arrhythmia detection. Ear-PPG allows for blood oxygenation and heart rate monitoring, particularly in cardiovascular applications (Ferlini et al., Reference Ferlini, Montanari, Min, Li, Sassi and Kawsar2021). Additionally, in-ear temperature sensing enables continuous core body temperature tracking for fever detection and wellness monitoring (Olson et al., Reference Olson, O’Brien, Lin, Fabry, Hanke and Schroeder2023). Despite this diversity, the growing interest in ear-EEG underscores its potential as a cornerstone of hearable-based neurotechnology, offering both clinical utility and everyday usability (Ne et al., Reference Ne, Muzaffar, Amlani and Bance2021).

3.1. Variance auditory meatus curvature

To accurately capture EEG signals, it is essential to position multiple EEG channels in close proximity to the patient’s auditory meatus. However, due to the inherent anatomical variability of patients’ auditory meatus, innovative solutions are required to address this challenge.

Various approaches have been explored in response to this challenge. A notable strategy involves utilizing 3D scanning technology to meticulously capture the precise dimensions of an individual’s auditory meatus. Subsequently, an electrode can be crafted with meticulous accuracy based on the acquired scanning data. Kappel et al.’s work has contributed to this area of research (Kappel et al., Reference Kappel, Rank, Toft, Andersen and Kidmose2018). However, a prevailing technique involves 3D printing personalized earpieces that incorporate pin electrodes placed at specific positions. With the continuous advancement of 3D printing technologies, electrodes can be seamlessly integrated into the earpiece alongside all required electrical components, resulting in a highly personalized and optimized solution.

Another pioneering endeavor in this domain has been undertaken by Wang et al., who introduced a novel methodology centered around a spiral electrode (Z Wang et al., Reference Wang, Shi, Zhang, Zheng, Li, Jiao, Cheng, Wang, Zhang, Chen, Chen, Wang, Xie, Wang, Ma, Gao and Feng2023). This electrode boasts a dual-layer design, encompassing SMPs containing an EAL and an EEGDL (Wang et al., Reference Wang, Shi, Zhang, Zheng, Li, Jiao, Cheng, Wang, Zhang, Chen, Chen, Wang, Xie, Wang, Ma, Gao and Feng2023). The EEGDL comprises Au wires and insulating polyimide (PI) (Wang et al., Reference Wang, Shi, Zhang, Zheng, Li, Jiao, Cheng, Wang, Zhang, Chen, Chen, Wang, Xie, Wang, Ma, Gao and Feng2023). A noteworthy feature of this electrode design is its capability to expand and spiral within the auditory meatus upon the application of heat (Wang et al., Reference Wang, Shi, Zhang, Zheng, Li, Jiao, Cheng, Wang, Zhang, Chen, Chen, Wang, Xie, Wang, Ma, Gao and Feng2023). This innovative approach enables the electrode to seamlessly adapt to the unique curvature of the auditory meatus during the expansion process, resulting in a customized fit tailored to the individual’s anatomical features. Kaveh et al.’s work contributed to this field by developing an ear-EEG electrode with an extendable plastic structure that effectively conforms its electrodes to the curvature of the ear (Kaveh et al., Reference Kaveh, Doong, Zhou, Schwendeman, Gopalan, Burghardt, Arias, Maharbiz and Muller2020).

The size of the ear canal can vary throughout different age groups. For example, for the age group 18–30 years, the right ear length is 60.65 $ \pm $ 4.71 mm, while for the age group 51–64 years, the right ear length is 65.05 $ \pm $ 3.57 mm (Japatti et al., Reference Japatti, Engineer, Reddy, Tiwari, Siddegowda and Hammannavar2018). Fourteen different ear shapes are being classified (Martinez et al., Reference Martinez, Hwee, Yap, De Chua, Kamath, Chung, Teo, Tan, Dritsas and Simpson2023). There are also gender differences being found where men’s first bend upper, lower, anterior, and posterior lengths for both ears are longer than those for women (Yu et al., Reference Yu, Lee, Wang, Chen, Fan, Peng, Tu, Chen and Lin2015). The separation of the human ear canal can be seen in Table 3.

Table 3. Percentage distribution of human ear dimensions, including length of the ear canal and area ranges, based on anthropometric data (Martinez et al., Reference Martinez, Hwee, Yap, De Chua, Kamath, Chung, Teo, Tan, Dritsas and Simpson2023)

Different approaches have been developed to address the difference in auditory. For example, 3D-printing technologies have been used to create personalized electrodes that ensure a proper fit, a concept that has long been applied in the hearing-aid field (Looney et al., Reference Looney, Park, Kidmose, Rank, Ungstrup, Rosenkranz and Mandic2011). Beyond individualized ear molds, mature solutions from the audiology and hearing-aid industries include the use of biocompatible silicone interfaces and soft polymer materials that can maintain comfort and reduce irritation during prolonged wear. Recent research prototypes have further explored dry and flexible electrodes, conductive coatings, and microstructured surfaces that improve skin contact without extensive preparation (Burgar, Reference Burgar2022). Looking ahead, future directions may involve self-adaptive ear interfaces, advanced impedance modeling for electrode design, and the translation of standardized manufacturing pipelines from hearing aids to hearables, which together could significantly improve signal fidelity and user acceptance.

3.2. Motion artifacts

Motion artifacts represent an ongoing challenge in the field of ear-EEG measurements. These artifacts, unwanted electrical signals induced by physical movements of the body or the measurement system, pose a significant issue for the accurate acquisition of EEG signals (Seok et al., Reference Seok, Lee, Kim, Cho and Kim2021). Unlike scalp EEG, ear-EEG is less affected by eye blinking but more susceptible to motion artifacts originating from jaw and head movements, primarily due to the extended wire connections involved (Kappel et al., Reference Kappel, Looney, Mandic and Kidmose2017). The magnitude of these motion artifacts is a particular concern, often exceeding that of the EEG signal by at least an order of magnitude (Seok et al., Reference Seok, Lee, Kim, Cho and Kim2021). The consequences of these artifacts are far-reaching and include a reduced accuracy in automated signal sequence classification for clinical diagnostic purposes and disruptions in the smooth operation of BCI systems (Seok et al., Reference Seok, Lee, Kim, Cho and Kim2021). Given the need for daily monitoring and real-world applications of hearable devices, effectively managing and mitigating motion artifacts is crucial to ensure the reliability and clinical utility of ear-EEG measurements. Existing mitigation strategies, such as accelerometer-based noise filtering, have shown moderate success (Occhipinti et al., Reference Occhipinti, Davies, Hammour and Mandic2022), but they struggle to differentiate true neural signals from muscle activity. Novel approaches, such as adaptive filtering using independent component analysis and deep learning-based artifact removal, have demonstrated promise (Phadikar et al., Reference Phadikar, Sinha and Ghosh2020), but their real-time applicability in wearable devices remains an open question due to computational constraints.

A study introduced an innovative ear-EEG electrode design that employs memory foam as its base material (Goverdovsky et al., Reference Goverdovsky, Rosenberg, Nakamura, Looney, Sharp, Papavassiliou, Morrell and Mandic2017). This choice of material effectively mitigates motion artifacts arising from both minor and substantial mechanical deformations of the ear canal walls (Goverdovsky et al., Reference Goverdovsky, Rosenberg, Nakamura, Looney, Sharp, Papavassiliou, Morrell and Mandic2017). This electrode design also incorporates using two microphones to capture acoustic signals traversing through dense head tissues and uses these signals to cancel the motion artifacts in the EEG signal (Goverdovsky et al., Reference Goverdovsky, Rosenberg, Nakamura, Looney, Sharp, Papavassiliou, Morrell and Mandic2017).

3.3. Design constraints and miniaturization challenges

The task of seamlessly integrating the complete analog front end and the MCU into a single earpiece to facilitate the wireless transmission of digital signals to a designated device remains a substantial challenge in this field. This challenge is further accentuated by the necessity of minimizing the distance between the electrode and the analog front end and the size of the auditory meatus (Mozaffari et al., Reference Mozaffari, Nash and Tucker2021). One suggested approach entails the strategic distribution of electrical components around the external ear rather than direct insertion into the ear canal (Kaveh et al., Reference Kaveh, Doong, Zhou, Schwendeman, Gopalan, Burghardt, Arias, Maharbiz and Muller2020). However, even with this approach, the visibility of these components remains noticeable, potentially causing social discomfort. In practical scenarios, this situation might require the use of additional accessories, such as a hat, to effectively conceal the components.

It is important to acknowledge that while addressing this challenge entails inherent complexities, its resolution is attainable. The utilization of ASICs has emerged as a promising solution; however, it is fitting to recognize that producing a limited quantity of ASICs can involve significant costs. As time progresses, the continued advancement of manufacturing and electronics technologies contributes to the ongoing miniaturization of electronic components. This enduring trend plays a pivotal role in facilitating the seamless integration of all necessary components within the confined space of the earpiece.

4.1. Multi-sensor hearables

The integration of multiple sensors into hearable devices enables the simultaneous acquisition of diverse physiological signals, transforming these devices into powerful platforms for continuous health monitoring and medical diagnostics. This multi-sensor approach broadens the scope of measurable parameters, enhances data fusion capabilities, and improves signal quality through redundancy and noise mitigation techniques. By applying machine learning and advanced signal processing, multi-sensor hearables can derive more meaningful insights from complex physiological data, supporting both consumer wellness and clinical applications. Additionally, such sensor integration holds promise for other domains, including brain-machine interfaces (Martin et al., Reference Martin, Boehler, Hollander, Kinney, Hitt, Kudva and Sugar2022; Bengler et al., Reference Bengler, Harbauer and Fleischer2023).

Żyliński et al. introduced a hearable device capable of monitoring ECG, PPG, and heart sounds, providing comprehensive and multimodal heart rate monitoring (Żyliński et al., Reference Żyliński, Nassibi, Occhipinti, Malik, Bermond, Davies and Mandic2024). Similarly, Montanari et al. developed a hearable integrating kinetic, acoustic, optical, and thermal sensors, leveraging machine learning for advanced data analysis (Montanari et al., Reference Montanari, Thangarajan, Al-Naimi, Ferlini, Liu, Balaji and Kawsar2024). These innovations highlight the potential of hearables as versatile platforms for real-time health monitoring.

Such advancements demonstrate the potential of multi-sensor systems to transform hearables into powerful tools for continuous health tracking, with applications spanning general wellness to clinical diagnostics.

4.2. Advanced signal processing techniques

Hearables are becoming increasingly sophisticated, integrating advanced signal processing techniques to improve signal quality, enhance user experience, and expand their range of applications. Effective signal processing is crucial for extracting meaningful physiological data, ensuring reliable real-time monitoring, and enabling applications such as health tracking, cognitive assessments, and biometric authentication.

One of the key challenges in hearables is enhancing the SNR to ensure high-quality physiological data acquisition. Techniques such as filtering to remove noise and artifacts, segmentation to isolate relevant events, and feature extraction to identify critical signal patterns are commonly employed. These approaches are particularly useful in capturing bioelectrical signals like EEG, ECG, and PPG, as well as acoustic and temperature signals. Due to the compact nature of hearables, designing efficient AFE circuits remains a challenge. To address this, integrated circuit AFE designs have been explored to optimize signal acquisition while minimizing size and power consumption (Xu et al., Reference Xu, Mitra, Van Hoof, Yazicioglu and Makinwa2017).

To overcome size and power constraints, several low-cost and efficient solutions have been developed. For instance, Teversham et al. designed a low-cost physiological monitoring system using an ESP32 MCU, integrating analog front-end components, such as a bandpass filter, differential amplifier, and notch filter (Teversham et al., Reference Teversham, Wong, Hsieh, Rapeaux, Troiani, Savolainen, Zhang, Maslik and Constandinou2022). This approach enables real-time signal decoding at a fraction of the cost of traditional medical devices, demonstrating the feasibility of low-cost, wearable physiological monitoring.

Utilizing external computational resources for additional signal processing offers another practical solution for enhancing portability and efficiency. Hölle and Bleichner introduced a system where a commercial amplifier and general-purpose hearable sensors were paired with a smartphone for real-time sound and biometric signal analysis (Hölle and Bleichner, Reference Hölle and Bleichner2023). By offloading computation to a smartphone, the hearable device itself can remain compact and energy-efficient while still benefiting from advanced signal processing capabilities.

Power efficiency is a critical factor for long-term hearable use. Lee et al. developed a low-power communication system utilizing body channel transmission, reducing power consumption and enabling continuous monitoring over extended periods (J Lee et al., Reference Lee, Lee, Ha, Kim, Lee, Gweon, Jang and Yoo2019). This type of innovation is essential for daily health monitoring applications, such as heart rate tracking, respiratory monitoring, and stress detection.

Dedicated chips designed for hearables focus on optimizing high input impedance and low noise to effectively capture bioelectrical and acoustic signals in real-world environments. For example, Zheng et al. introduced a time-division multiplexing-based AFE with a capacitively coupled chopper design, achieving superior noise performance (Zheng, Reference Zheng2024). Similarly, Jin et al. developed a low-power AFE ASIC tailored for wearable applications, ensuring minimal power consumption while maintaining signal integrity (Jin et al., Reference Jin, Hu, Zhao, Jiang, Ye, Wang and Qin2023).

These advancements in analog front-end design, power efficiency, and AI-driven processing are transforming hearables into powerful, multifunctional health monitoring tools. By addressing challenges related to cost, size, and signal processing, hearables are expanding their role in applications ranging from cardiovascular health monitoring and sleep tracking to cognitive enhancement and biometric security. It is important to note that Apple released a patent in the ear EEG domain, which enables the auto-selection of the best electrode for the current situation (Azemi et al., Reference Azemi, Moin, Pragada, Lu, Powell, Minxha and Hotelling2023).

4.3. Deployment of AI models

Advances in AI and edge computing hardware have significantly enhanced the feasibility of deploying AI models on wearable devices. These developments offer substantial benefits, including the delivery of actionable insights, early detection of health issues, and the prioritization of critical data for healthcare professionals. Furthermore, on-device training enables the creation of personalized models tailored to the unique physiological characteristics of individual users. However, due to the inherent constraints of hearables, such as limited computational resources and reduced input data channels, AI models must be carefully optimized for efficiency.

In the domain of EEG signal processing, several studies have demonstrated the promise of edge-optimized AI models. For example, Contreras et al. proposed an ultra-low-power seizure detection model specifically designed for integration into wearable systems, positioning it as a strong candidate for next-generation hearables (Herbozo Contreras et al., Reference Herbozo Contreras, Huang, Yu, Nikpour and Kavehei2024). Similarly, Zhu et al. introduced a lightweight edge model for sleep staging based on single-channel EEG data, making it particularly suitable for wearable applications (Zhu et al., Reference Zhu, Wang, He and Zhang2022). Complementing these approaches, Huang et al. developed a dedicated chip architecture for deploying AI models efficiently on edge devices (Huang et al., Reference Huang, Wang, Ho, He and Fang2019).

The integration of AI significantly enhances the capabilities of hearables by improving signal interpretation, automation, and user interactivity. For instance, Mai et al. implemented a real-time emotion classification system using a 1D-Convolutional Neural Network on an embedded platform, utilizing fast Fourier transform and power spectral density features (Mai et al., Reference Mai, Nguyen and Chung2023). AI-driven methods can also improve data quality by reducing noise and artifacts. Jayas et al. demonstrated an algorithm capable of detecting motion and muscle artifacts in physiological signals, thereby enhancing the reliability of downstream analysis (Jayas et al., Reference Jayas, Adarsh, Muralidharan, Gubbi and Pal2023).

Beyond signal enhancement, AI integration empowers hearables to deliver autonomous health insights in real time, often without requiring professional oversight. Neuromorphic computing, in particular, enables ultra-low-power, continuous on-device processing, which is ideal for long-term health monitoring (Sharifshazileh et al., Reference Sharifshazileh, Burelo, Sarnthein and Indiveri2021). Similarly, Field Programmable Gate Arrays (FPGAs) have been leveraged for energy-efficient AI inference. For example, Al-Ashmouny et al. showcased FPGA-based processing for sleep apnea detection using physiological data, underscoring the potential for real-time assessments in wearable applications (Al-Ashmouny et al., Reference Al-Ashmouny, Hamed and Morsy2006).

For ECG signals, multiple efforts have aimed at achieving high performance within resource-constrained environments. Huang et al. developed an edge-friendly ECG classification model, while Huang et al. proposed a preprocessing-free model optimized for detecting heart abnormalities on low-power devices (Huang et al., Reference Huang, Herbozo Contreras, Leung, Yu, Truong, Nikpour and Kavehei2024a; Huang et al., Reference Huang, Herbozo Contreras, Yu, Truong, Nikpour and Kavehei2024b). Additionally, Huang et al. presented a novel framework for training large-scale models directly on edge platforms (Huang et al., Reference Huang, Yu, Contreras, Eshraghian, Truong, Nikpour and Kavehei2025). In a practical demonstration, Meza-Rodriguez et al. successfully deployed a heart abnormality classification system on MCU, highlighting the viability of implementing sophisticated AI on compact hardware (Meza-Rodriguez et al., Reference Meza-Rodriguez, De La Cruz and Cáceres-DelAguila2023).

Collectively, these efforts highlight the transformative potential of AI-enabled wearables. By supporting real-time analysis, personalization, and improved accessibility, AI-driven hearables are set to revolutionize both clinical and consumer health monitoring, positioning them as indispensable tools in modern healthcare.

4.4. Advanced material technologies

Advancements in material technologies have enabled the development of electrodes and components that enhance adaptability and comfort, allowing for a more personalized fit within the ear canal. These innovations improve both the functionality and wearability of in-ear biosignal monitoring systems, addressing challenges such as signal quality, durability, and long-term usability.

Ahn et al. introduced a technology utilizing 3D printing with silver ink to create stretchable electrodes that can be directly applied to the skin for biosignal monitoring (Ahn et al., Reference Ahn, Duoss, Motala, Guo, Park, Xiong, Yoon, Nuzzo, Rogers and Lewis2009). Similarly, Kumar et al. developed a flexible supercapacitor capable of serving as a power source for wearable devices, while Xu et al. designed a stretchable battery that supports wireless charging, further enhancing the feasibility of fully flexible wearable systems (Xu et al., Reference Xu, Zhang, Cho, Lee, Huang, Jia, Fan, Su, Su, Zhang, Cheng, Lu, Yu, Chuang, Kim, Song, Shigeta, Kang, Dagdeviren, Petrov, Braun, Huang, Paik and Rogers2013; Kumar et al., Reference Kumar, Di Mauro, Zhang, Pezzella, Soavi, Santato and Cicoira2016).

Hydrogel-based electrodes also offer unique advantages, particularly in the development of semi-dry electrodes. These materials can achieve low impedance between the electrode and the skin while maintaining durability and long-term stability. Ge et al. developed an EEG electrode incorporating hydrogels to achieve both low impedance and high longevity (Ge et al., Reference Ge, Guo, Gong, Han, Feng, Ji, Sun, Gao, Bian and Xu2023). However, the effectiveness of generalized ear-EEG electrodes may vary significantly due to the use of elastic materials. Since these electrodes rely on the tension of plastic components, variations in material elasticity can impact electrode-skin contact, potentially leading to inconsistencies in impedance and signal quality.

These advancements collectively pave the way for the development of fully flexible, high-performance in-ear biosignal monitoring systems. By improving comfort, durability, and energy efficiency, these materials enhance the practicality of hearables for long-term health monitoring and medical applications.

4.5. Advanced manufacturing technology

Recent advancements in manufacturing technology have significantly improved the efficiency and feasibility of producing complex, highly specialized hearable devices. Techniques such as 3D printing have revolutionized the production process, enabling precise customization to accommodate individual anatomical and functional requirements (Persad and Rocke, Reference Persad and Rocke2022).

For instance, 3D printing allows for the fabrication of personalized ear canal components, enhancing both comfort and signal quality by ensuring an optimal fit (Kaveh et al., Reference Kaveh, Doong, Zhou, Schwendeman, Gopalan, Burghardt, Arias, Maharbiz and Muller2020). Additionally, advanced material technologies, such as biocompatible polymers and conductive inks, facilitate the seamless integration of sensors and electronic components directly into the device structure. This integration reduces assembly complexity, improves signal fidelity, and enhances device durability.

Furthermore, miniaturization techniques have enabled the development of compact, low-profile hearables capable of housing multifunctional sensors, wireless communication modules, and energy-efficient processing units. These innovations help address challenges related to anatomical variability, device miniaturization, and long-term wearability, paving the way for the next generation of customized, high-performance hearable devices.

4.6. Brain–computer interface

As portable devices for detecting EEG signals, hearables have the potential to significantly advance BCI technology by providing a discreet, wearable alternative to traditional EEG systems. BCIs enable direct communication between the brain and external devices, with applications ranging from neuroprosthetic control and cognitive monitoring to assistive communication for individuals with paralysis.

Current BCI systems typically rely on either invasive electrodes, such as electrocorticography implants, or noninvasive scalp EEG systems that use wet electrodes and multichannel headsets (Miller et al., Reference Miller, Hermes and Staff2020; Hsieh et al., Reference Hsieh, Alawieh, JdR and Wang2024). Although scalp EEG provides higher spatial resolution than ear-EEG, its bulky setup, dependence on gel-based electrodes, and susceptibility to motion artifacts greatly limit its practicality outside controlled laboratory settings (Asayesh et al., Reference Asayesh, Warsito, Haueisen, Fiedler and Vanhatalo2025). In contrast, ear-EEG offers a more portable and user-friendly alternative, and has been increasingly recognized as a promising solution for assisting individuals with disabilities in controlling prosthetic devices (Maibam et al., Reference Maibam, Pei, Olikkal, Vinjamuri and Kakoty2024; Liu et al., Reference Liu, Wang, Liu, Chen, Pereira, Doda, Ke, Wang, Wen and Tong2025).

With the development of high-quality ear-EEG recording, hearables could address these limitations by offering:

1. 1. A more user-friendly, noninvasive interface for BCI applications.

2. 2. Continuous brain monitoring in real-world environments.

3. 3. Improved social acceptability, as they resemble common earbuds rather than medical devices.

However, in order to achieve the use of BCI, there are still some specific challenges that remain to be solved:

1. 1. Low spatial resolution: Ear-EEG captures fewer electrodes and less choice in positions than traditional scalp EEG, limiting the amount of brain activity that can be recorded with high signal quality (Looney et al., Reference Looney, Kidmose and Mandic2014). As a result, certain neural signals may not be detected, reducing the effectiveness of BCIs. Future advancements may include the development of multi-electrode arrays or signal reconstruction techniques to enhance spatial resolution and improve overall performance.

2. 2. Processing constraints: BCI applications require low-latency signal processing to enable real-time interactions. However, the computational capacity of hearables is constrained by size, thermal limits, and power efficiency (Haider and Guragain, Reference Haider and Guragain2023). Lightweight edge-computing architectures, neuromorphic processors, or dedicated BCI hardware accelerators could provide solutions to enhance real-time processing capabilities without excessive power consumption.

3. 3. Wireless data transmission: Real-time BCI applications demand high-bandwidth, low-latency data transfer, which is challenging given the current wireless communication standards available in hearables. Bluetooth low energy offers power efficiency but is limited in data throughput, while body channel communication presents an emerging alternative with the potential to reduce interference and improve transmission reliability (Debener et al., Reference Debener, Kranczioch and De Vos2016). Future work should focus on hybrid communication methods that optimize between latency, power consumption, and bandwidth availability.

4. 4. Power constraints: The computational demands of real-time EEG processing, multichannel high sampling rate acquisition, and continuous wireless transmission significantly increase power consumption, posing a major challenge for portable BCIs (Edelman et al., Reference Edelman, Zhang, Schalk, Brunner, Müller-Putz, Guan and He2024). Current battery technology is insufficient for long-duration operation without frequent recharging. Advancements in energy-efficient circuits, ultra-low-power AI models, and energy harvesting technologies (e.g., thermoelectric or piezoelectric power generation) are needed to enhance the viability of long-term, real-world use.