Structure Assisted Spectrum Sensing for Low-power Acoustic Event Detection Nakul Garg, Harshvardhan Takawale, Yang Bai, Irtaza Shahid, Nirupam Roy {nakul22, htakawal, yangbai8, irtaza, niruroy}@umd.edu University of Maryland College Park ABSTRACT Acoustic sensing has conventionally been dependent on high- frequency sampling of analog signals and frequency domain analysis in digital domain which is power-hungry. While these techniques work well for regular devices, low-power acoustic sensors demand for an alternative approach. In this work, we propose Lyra, a novel low-power acoustic sensing architecture that employs carefully designed passive struc- tures to filter incoming sound waves and extract their fre- quency components. We eliminate power-hungry compo- nents such as ADC and digital FFT operations and instead propose to use low-power analog circuitry to process the signals. Lyra aims to provide a low-power platform for a range of maintenance-free acoustic event monitoring and ambient computing applications. CCS CONCEPTS • Computer systems organization → Sensor networks; Embedded systems; Sensors and actuators. KEYWORDS Low-power sensing; IoT; Ambient sensing; Ambient com- puting; Acoustic metamaterial; Spectral sensing; Structural filters; passive computing ACM Reference Format: Nakul Garg, Harshvardhan Takawale, Yang Bai, Irtaza Shahid, Niru- pam Roy. 2023. Structure Assisted Spectrum Sensing for Low-power Acoustic Event Detection. In Cyber-Physical Systems and Internet of Things Week 2023 (CPS-IoT Week Workshops ’23), May 09–12, 2023, San Antonio, TX, USA. ACM, New York, NY, USA, 7 pages. https://doi.org/10.1145/3576914.3589562 1 INTRODUCTION Acoustic sensing has emerged as a key technology in ambient computing, enabling devices to understand and respond to This work is licensed under a Creative Commons Attribution International 4.0 License. CPS-IoT Week Workshops ’23, May 09–12, 2023, San Antonio, TX, USA © 2023 Copyright held by the owner/author(s). ACM ISBN 979-8-4007-0049-1/23/05. https://doi.org/10.1145/3576914.3589562 Traditional approach: Lyra approach: Sampling FFT Structural filters with microphones Fast Fourier TransformADCEvent Microphone Fourier coefficients Frequency detection triggers Low-power analog comparatorsEvent Classification Classification Figure 1: Traditional approach vs Lyra approach for spectral sensing. Lyra replaces ADC and FFT using structural filters and analog comparators. their environment by detecting and analyzing events through sound. Furthermore, the ubiquitous nature of sound hasmade acoustic sensing and communication more accessible, with a majority of smart devices equipped with built-in micro- phones. Acoustic signals have demonstrated their utility in a diverse range of applications, including but not limited to gesture recognition and tracking [17, 30, 36, 48], early medical diagnosis [35], source localization [19, 43], acoustic imaging [3, 31], voice interfaces [5, 9, 41], drone defense and tracking [20, 32] and finer-grained activities such as eye blinks and heartbeats [29, 44]. The infrastructure for the future of ambient computing and maintenance-free sensing are in continuous pursuit of low- power sampling and processing methods. Like other modali- ties, acoustic sensing systems depend on sampling of analog signals and frequency domain analysis. These two opera- tions are considered bottleneck in ultra-low-power sensing architecture. Fourier transforms are a crucial tool for spectral analysis and are extensively utilized in acoustic communica- tion and sensing systems. Typically, acoustic systems first sample an incoming signal at Nyquist rate [26], a rate of at least twice the frequency of the signal, to acquire a digi- tal copy of the signal, after which a Fast Fourier Transform (FFT) algorithm is employed to obtain the frequency domain representation of the signal. This high sampling rate and fourier transform operation consume a significant amount of 278 https://doi.org/10.1145/3576914.3589562 https://creativecommons.org/licenses/by/4.0/ https://creativecommons.org/licenses/by/4.0/ https://doi.org/10.1145/3576914.3589562 http://crossmark.crossref.org/dialog/?doi=10.1145%2F3576914.3589562&domain=pdf&date_stamp=2023-05-09 CPS-IoT Week Workshops ’23, May 09–12, 2023, San Antonio, TX, USA Nakul Garg, Harshvardhan Takawale, Yang Bai, Irtaza Shahid, Nirupam Roy energy. The reliance on such high energy-consuming opera- tions presents a significant challenge for low-power ambient acoustic sensors that must continuously listen for acoustic cues without depleting their battery life. In this paper, we propose Lyra1, a novel approach that utilizes passive structures for low-power acoustic frequency detec- tion. As shown in Figure 1, Lyra gets rid of the power-hungry components of conventional systems, such as ADC and FFT, and replaces them with passive structures and low-power analog comparators. The passive structures are narrowband acoustic filters that are sensitive to a specific resonant fre- quency. Multiple of these structures are arranged in parallel to create a unique block that is highly sensitive to desired set of frequencies. These structures are incorporated with microphones that capture the resonating acoustic signal, and a low-power analog comparator is employed to detect if the energy is above a predefined threshold, indicating the pres- ence of the resonant frequency. This approach significantly reduces the power consumption of the sensor and enables energy harvesting sensors which have limited energy bud- gets to sense acoustic signals and function continuously in remote areas for extended periods of time. The low-power design addresses the issue of high energy consumption by conventional systems and allows low-power devices to continuously monitor acoustic signals while mini- mizing energy consumption, which opens possibilities for new applications. Lyra can enable the deployment of en- ergy harvesting sensors in remote areas such as forests, for wildlife monitoring and tracking, and as SOS devices, always listening for emergency signals played by lost individuals seeking help. The sensors would require no maintenance and offer a cost-effective and long-term solution for various applications, including environmental monitoring, disaster response, and other remote sensing applications. Lyra’s architecture is designed to extract a time-series rep- resentation of the frequencies present in incoming sound waves. Our passive structures are designed based on the principles of standing wave resonance [33] and Helmhotz resonance [46]. The placement of microphones inside the res- onating structure is carefully determined to create a salient feature for resonance detection. We then use passive en- velope detectors, low-power analog comparators and logic gates to classify the occurrence of resonance in the structure, thus indicating the presence of the resonating frequency. Needless to say, this paper is only a first step toward this broader vision, and we aim to inquire about the opportunities it presents. We start by experimenting with the technique of 1Named after the Lyra constellation in the northern sky, representing a harp which is a musical instrument with strings that produce standing waves. acoustic event detection system that employs passive struc- tures to identify frequencies present in an acoustic signal. 2 APPLICATION SCENARIOS This paper focuses on exploring the technical cores that en- able structure-assisted spectrum sensing. While the ideal applications and application-specific advantages and limita- tions are yet to be analyzed, we outline a few target applica- tion domains here. Ultra-low-power wake-word detection: The core idea of Lyra can lead to applications that require frequency anal- ysis. One specific example is wake word detection, which recognizes a specific word or phrase spoken by the user using frequency analysis of human speech. Some common examples of wake words include "Hey Siri" for Apple’s vir- tual assistant, "OK Google" for Google Assistant, and "Alexa" for Amazon’s voice assistant. We look forward to applying Lyra on more applications requiring ultra-low-power and maintenance-free acoustic signal processing. Large scale ambient computing: The inexpensive design of Lyra can be seamlessly integrated into infrastructure and enable acoustic interfaces for device-free event detection and thereby enabling closer interaction between computing infrastructure and the living space, both indoor and outdoor. For example, smart city infrastructure can use Lyra to detect the number of cars passing through and adjust traffic signals to optimize traffic flow. Inaccessible locations can automate monitoring, such as an elevator shaft can monitor its long- term structural consistency using acoustic cues. Maintenance-free SOS system: With the features of ultra- low-power and maintenance-free, this work can be a means for SOS system that can be widely deployed in a variety of situations, such as hiking, boating, or traveling in remote areas where access to emergency services is limited. It can be used by people who may require urgent assistance due to medical emergencies, accidents, or other unforeseen events. Wildlife monitoring: It is natural to wonder if Lyra can be an interface for acoustic monitoring of wildlife – a scenario where scheduled maintenance is difficult and standard power supply is limited. Since the proposed acoustic interface needs limited power and maintenance, it can be a critical tool to better understand the ecology and behavior of wildlife pop- ulations and make informed decisions about how best to protect them. 3 THEORETICAL FOUNDATIONS 3.1 Passive filtering An acoustic passive filter is a type of filter used in audio sys- tems to eliminate or attenuate specific frequencies. Passive 279 Structure Assisted Spectrum Sensing for Low-power Acoustic Event Detection CPS-IoT Week Workshops ’23, May 09–12, 2023, San Antonio, TX, USA Figure 2: The cochlea located in the inner ear is a natural multiresonant structure that provides acoustic filtering. filters are designed to workwithout an external power source and are made up of passive components. Interestingly, the use of passive acoustic filtering is pretty common in nature. For instance, the structure of the human ear plays a crucial role in shaping the frequency response of sounds that reach the eardrum. The shape and size of the external ear create a frequency-dependent filtering effect on incoming sounds. This filtering effect is due to the fact that different frequen- cies of sound waves interact differently with the shape and size of the external ear [42]. In particular, the ear acts as a frequency-dependent amplifier, boosting sounds in certain frequency ranges while attenuating others. This amplifica- tion effect is most pronounced in the range of frequencies that are important for speech perception. Moreover, cochlea in the inner ear also acts as a passive filter to help us hear and perceive different frequencies of sound [38], as shown in Figure 2. The cochlea is a spiral-shaped organ located in the inner ear that is responsible for converting sound waves into neural signals that are sent to the brain for processing. Unlike using frequency domain transforms (e.g., FFT) for analysis used in modern acoustic systems, cochlea senses frequencies through the vibration on different parts of it. Specifically, high-frequency sounds cause vibrations near the base of the cochlea, while low-frequency sounds cause vibrations near the apex. 3.2 Standing wave resonance In the discussion of the impact of structure on sound waves, the understanding of standing waves is relevant. A standing wave is a type of wave pattern that occurs when two waves of the same frequency and amplitude travel in opposite di- rections and interfere with each other [33]. When the waves interact with each other, the resulting wave appears to "stand still" in place and does not move. As shown in Figure 3a, we can achieve this using a pipe with one closed end that can be utilized to generate a reflected signal and thus, create a standing wave. The shape and size of the medium or a tube defines the properties of a standing wave for a given frequency. The mathematical equation of a standing wave is Nodes Antinodes (a) Volume of cavity Neck (b) Figure 3: Lyra’s structural acoustic filters. (a) A longitudi- nal single-frequency resonant pipe (b) A spherical multi- frequency Helmholtz resonator. given by 𝑦 (𝑥, 𝑡) = 𝑠𝑖𝑛(2𝜋𝑥/_)𝑐𝑜𝑠 (2𝜋 𝑓 𝑡) (1) where 𝑥 is the spatial distance, _ is the wavelength, and 𝑓 is the frequency of the wave. The term 𝑠𝑖𝑛(2𝜋𝑥/_) repre- sents the sinusoidal shape in space, and the term 𝑐𝑜𝑠 (2𝜋 𝑓 𝑡) describes the up-down oscillatory motion in time domain. Standing waves also explain the production of sound by mu- sical instruments. Instruments create standing waves on a string or in a tube or pipe and the perceived pitch of the sound is related to the frequency of the standing wave. 3.3 Helmholtz resonance Helmholtz resonators amplify certain frequencies when os- cillating air pressures meet cavities on their way [46]. As shown in Figure 3b, a Helmholtz resonator consists of a cav- ity of air connected to the outside world by a narrow neck or tube. When sound waves enter the container, they set up a standing wave pattern that resonates at a specific frequency determined by the volume of the container and the size of the opening [1]. The resonance frequency 𝑓 is 𝑓 = 𝑐 2𝜋 √︂ 𝑆 𝑉𝐿 (2) where 𝑐 is the speed of sound, 𝑆 is the cross-sectional area of the tube,𝑉 is the volume of the cavity, and 𝐿 is the length of the tube. 4 LYRA ARCHITECTURE Our objective is to design an architecture using passive struc- tures to extract a time-series representation of the frequen- cies present in an acoustic signal, similar to a spectrogram. To accomplish this, we designed a series of structural filters and a low-power analog logic to identify the presence of fre- quencies. A standing wave is an interference phenomenon of waves characterized by nodes and antinodes. As illustrated in Figure 3a, nodes are points on the standing wave that re- main stationary with no pressure variations, while antinodes are points that undergo maximum pressure variations. We aim to detect these nodes and antinodes inside the structure, whose precise location is determined by the wavelength of the incoming acoustic wave. 280 CPS-IoT Week Workshops ’23, May 09–12, 2023, San Antonio, TX, USA Nakul Garg, Harshvardhan Takawale, Yang Bai, Irtaza Shahid, Nirupam Roy To resonate at a specific frequency, we adjust the shape and size of the structure and place two microphones at a node and antinode, respectively. Unlike traditional methods that require sampling of the signals, we aim to employ a passive envelope detector followed by analog comparators to identify the energy threshold and categorize a node and antinode. A logical AND gate can classify the presence of both node and antinode, which indicates the detection of the frequency as a binary signal output, where 1 represents the frequency’s presence and 0 indicates its absence. Our approach of using passive structure and analog compo- nents effectively eliminates the need for energy-intensive ADC and FFT operations. Conventional ADCs and digital computations require a significant amount of power and can consume milli-watts of power [19, 45] whereas high impedance analog comparators [2] consume micro-watts of power (𝑝𝑜𝑤𝑒𝑟 ∝ 𝑣𝑜𝑙𝑡𝑎𝑔𝑒2/𝑖𝑚𝑝𝑒𝑑𝑎𝑛𝑐𝑒) and envelope detec- tors operate on passive electric components, resulting in zero additional energy consumption. 5 FEASIBILITY STUDY For initial testing, we conduct a simulation study using pas- sive structures utilizing both helmholtz and standing wave resonance. Figure 4 shows the spectral response of three helmholtz filters used in this simulation, where the filters are hollow cavities with resonant frequencies 800, 900, and 1300Hz. The standing wave structure is tuned to resonate at 2300Hz and Figure 5 shows the spectral responses of the sen- sors placed inside the structure. We record the signal using virtual sensors placed at the antinode and node locations of the standing wave. The shaded region shows the resonant frequency where the antinode sensor receives an amplified signal whereas the node sensor receives negligible power of signal, thus indicating a salient feature for resonance detec- tion. We conduct our simulation using Matlab and the k-wave acoustic toolbox, which is an acoustic wave propaga- tion model that accounts for reflections, diffraction, non- linearities and power law absorption for time domain anal- ysis. We perform a 2D simulation with an arbitrary signal containing multiple frequencies and record the response from virtual sensors. We then emulate the operation of a passive envelope detector by applying a squaring operation and a low-pass filter. Finally, we use a threshold-based de- tection to classify if a frequency is present or not. Figure 6 shows the output of the combined acoustic filters when the corresponding signal is transmitted. We observe that the series of helmholtz filters act like traditional frequency bin detectors and give a similar-looking spectrogram without actually sampling the signal or performing FFT. 1000 2000 3000 4000 Frequency (Hz) 10-3 10-2 10-1 100 101 Po w er Filter 1 Filter 2 Filter 3 Figure 4: Spectral response of three different helmholtz resonators with different harmonic frequencies. These filters are used as passive filters to detect presence of frequencies in an acoustic signal. 1000 2000 3000 4000 5000 Frequency (Hz) 10-4 10-2 100 Po w er Mic at Antinode Mic at Node Figure 5: Spectral response of the standing wave structure for two different microphones placed in the structure. One microphone is located at a node and another is located at an antinode of the standing wave. This creates a salient feature for the passive structure to detect a resonating frequency. To assess the feasibility of our proposed approach in the real- world, we fabricate the standing wave structure using 3D printing and conduct experiments to detect the resonating frequency. Figure 7 shows our experimental setup, where an acoustic signal is transmitted using a speaker towards the structure, which is designed to resonate at the fifth harmonic with frequency 𝑓 = 4200𝐻𝑧 and wavelength _ = 8.16𝑐𝑚. We place two microphones in the structure at the antinode and node locations of standing wave to extract the features for the standing wave. Figure 8 shows the time domain response of the two micro- phones, for input frequencies ranging from 200𝐻𝑧 to 6200𝐻𝑧. The figure shows that only at the resonance frequency the 281 Structure Assisted Spectrum Sensing for Low-power Acoustic Event Detection CPS-IoT Week Workshops ’23, May 09–12, 2023, San Antonio, TX, USA 0 0.5 1 1.5 2 Fr eq (k H z) 0 0.5 1 1.5 2 Fr eq (k H z) 0 200 400 600 800 1000 Time (ms) 0 0.5 1 1.5 2 Fr eq (k H z) 0 0.5 1 1.5 2 Fr eq (k H z) Transmitted signal Filter 1 Filter 2 Filter 3 Figure 6: Spectrogram of filtered responses from each of the helmholtz filters for a wide-band transmit signal. Mic A Mic B Distance = 30cm Speaker Structural filter Figure 7: One of the setups for the feasibility experiments. two microphones exhibit a high-energy contrast, high am- plitude for the microphone positioned at the antinode, while zero amplitude for the one at node, indicating successful detection. 6 LIMITATION AND FUTUREWORK Lyra is an initial exploration of utilizing passive structures for acoustic event detection and that there is room for further improvements and future work. Temperature variations: The speed of sound is affected by temperature variations in the medium, which in turn, af- fects the wavelength of the acoustic signal. Since the passive 4200Hz200Hz 1200Hz 2200Hz 3200Hz 5200Hz 6200Hz Antinode Node Resonant frequency Mic A Mic B Figure 8: The structure resonates to create a standing wave at 4200Hz (the 5𝑡ℎ harmonic. Plot shows the time domain amplitudes at two locations. structures are designed to detect wavelengths rather than frequencies from the principles of resonance, the system may fail to detect a frequency if temperature changes. To ac- commodate temperature-induced changes, one could employ reconfigurable structures. Such structures would enable a one-time actuation/reconfiguration to adapt to temperature variations by changing shape or size of the structure. Structure size: At lower frequencies, the physical size of structures can become considerably large. This is due to the slower speed of sound, which results in longer wavelengths. For example, an acoustic signal of 100 Hz has a wavelength of 3.43 meters in air. We aim to investigate novel miniaturized designs that leverage interferometry and metamaterials to enable the detection of longer wavelengths. 7 RELATEDWORK The core intuition of Lyra builds on the foundational works in physics and computational acoustics, while presenting a new sensing architecture with a unique goal to achieve ultra- low-power acoustic perception. We sample from relevant works in these fields. Structural acoustics:As sound waves pass through a cavity, they can be influenced by the structure of the cavity, causing certain frequencies to be amplified or suppressed. The art of using structure to manipulate sound waves is known for a very long time and its application is seen in many ancient architectures [50]. Previous works [12, 23, 40] have estab- lished theories for noise reduction using carefully designed acoustic structures, which have been improved upon by in- corporating sub-chamber structures [40], varying inlet and outlet sizes [12], or perforated liners [10]. Researchers have also investigated the use of acoustic metamaterials to de- sign structural materials that provide frequency-selective characteristics [52]. Xie et al [49] demonstrate that carefully 282 CPS-IoT Week Workshops ’23, May 09–12, 2023, San Antonio, TX, USA Nakul Garg, Harshvardhan Takawale, Yang Bai, Irtaza Shahid, Nirupam Roy designed acoustic structures can be used for selective listen- ing to solve cocktail party problems. Acoustic Voxels [28], proposed a method that assembles pre-designed structures into a complex geometry to achieve the desired acoustic filtering. Owlet [18, 19], showed the potential of using acous- tic microstructure to embed directional clues to the signal recorded by a microphone, enabling direction of arrival esti- mation without using large microphone arrays. One of our past works, SPiDR [3, 4], proposed an acoustic microstruc- ture to produce a cross-sectional map of the field of view for ultra-low-power spatial sensing. In this paper, we are build- ing on previous research and investigating the possibility of using acoustic structures for low-power spectrum sensing. Metamaterial resonators: Helmholtz resonators have been studied extensively in the field of acoustics and have been used in various applications, including noise reduction, sound absorption, and musical instruments [14]. The seminal work on Helmholtz resonators dates back to 19𝑡ℎ century by German physicist Hermann von Helmholtz [47]. Studies have explored the use of Helmholtz resonators in automo- tive exhaust systems to reduce engine noise [34], as well as in industry for damping heavy machinery noise [6, 51]. Helmholtz resonators have also been used in architectural acoustics to control sound reflections in large rooms and concert halls [24, 25]. Recent research has focused on the use of Helmholtz resonators in acoustic metamaterial for a perfect absorption of sound wave [39]. In summary, the study of Helmholtz resonators is an active area of research in acoustics for the past few decades. In this paper, we are exploring the concepts of Helmholtz resonators in designing passive sensors for acoustic spectrum sensing. Coded sensing in other modalities: Coded sensing is a technique used for data acquisition that involves taking mea- surements of a signal through a linear combination of a random or structured pattern. It has gained significant at- tention in the field of computer vision [11]. Previous work has explored the use of coded phase plates to capture high- resolution depth information [13], and all-focus images [8]. Levin et al [27] inserted a patterned occluder within the aperture of the camera lens, which enabled simultaneous recovery of both high-resolution depth and all-focus images. Additionally, coded exposure in the temporal domain has been used for motion deblurring [37]. Coded aperture imag- ing methods have also been employed in astronomy and medical imaging to capture the high-quality images [7, 16]. Similarly, the projection of structured light has been explored for 3D surface imaging [21]. In addition, the use of light mask- ing patterns has been proposed [15] and leveraged [22] for single-pixel imaging. In this paper, we investigate the poten- tial of using coded structures to develop low-power spectrum sensing techniques in the acoustic domain. 8 CONCLUSION This paper presents the feasibility experiments and early results of a new low-power acoustic sensing architecture. It shows the possibility of using carefully designed passive shapes for filtering incoming sound waves to reveal its fre- quency components. It then proposes a low-power sensing circuitry to use this filtered signal as a spectral feature vector for further processing. This sensing model aims to provide a low-power platform for a range of maintenance-free acoustic event monitoring and ambient computing applications. REFERENCES [1] M Alster. 1972. Improved calculation of resonant frequencies of Helmholtz resonators. Journal of sound and vibration 24, 1 (1972), 63–85. [2] Analog. 2023. https://www.analog.com/media/en/technical- documentation/data-sheets/MAX951-MAX954.pdf. [3] Yang Bai, Nakul Garg, and Nirupam Roy. 2022. Spidr: Ultra-low-power acoustic spatial sensing for micro-robot navigation. In Proceedings of the 20th Annual International Conference on Mobile Systems, Applica- tions and Services. 99–113. [4] Yang Bai, Nakul Garg, and Nirupam Roy. 2022. Ultra-low-power acous- tic imaging. In Proceedings of the 20th Annual International Conference on Mobile Systems, Applications and Services. 523–524. [5] Yang Bai, Nakul Garg, Harshvardhan Takawale, Anupam Das, and Nirupam Roy. 2023. Natural Voice Interface for the Next Generation of Smart Spaces. In Proceedings of the 24th International Workshop on Mobile Computing Systems and Applications. 140–140. [6] Valter Bellucci, Peter Flohr, Christian Oliver Paschereit, and Fulvio Magni. 2004. On the use of Helmholtz resonators for damping acoustic pulsations in industrial gas turbines. J. Eng. Gas Turbines Power 126, 2 (2004), 271–275. [7] Ezio Caroli, JB Stephen, G Di Cocco, L Natalucci, and A Spizzichino. 1987. Coded aperture imaging in X-and gamma-ray astronomy. Space Science Reviews 45 (1987), 349–403. [8] W Thomas Cathey and Edward R Dowski. 2002. New paradigm for imaging systems. Applied optics 41, 29 (2002), 6080–6092. [9] Ishan Chatterjee, Maruchi Kim, Vivek Jayaram, Shyamnath Gollakota, Ira Kemelmacher, Shwetak Patel, and Steven M Seitz. 2022. ClearBuds: wireless binaural earbuds for learning-based speech enhancement. In Proceedings of the 20th Annual International Conference on Mobile Systems, Applications and Services. 384–396. [10] Min-Chie Chiu. 2010. Shape optimization of multi-chamber mufflers with plug-inlet tube on a venting process by genetic algorithms. Ap- plied Acoustics 71, 6 (2010), 495–505. [11] Michał J Cieślak, Kelum AA Gamage, and Robert Glover. 2016. Coded- aperture imaging systems: Past, present and future development–A review. Radiation Measurements 92 (2016), 59–71. [12] Key Fonseca De Lima, Arcanjo Lenzi, and Renato Barbieri. 2011. The study of reactive silencers by shape and parametric optimization tech- niques. Applied Acoustics 72, 4 (2011), 142–150. [13] Edward R Dowski and W Thomas Cathey. 1994. Single-lens single- image incoherent passive-ranging systems. Applied Optics 33, 29 (1994), 6762–6773. [14] dsa2gamba and Abbottds. [n. d.]. Helmholtz resonators. https: //pressbooks.pub/sound/chapter/helmholtz-resonators/. Last accessed 12 March 2023. [15] Marco F Duarte, Mark A Davenport, Dharmpal Takhar, Jason N Laska, Ting Sun, Kevin F Kelly, and Richard G Baraniuk. 2008. Single-pixel 283 https://www.analog.com/media/en/technical-documentation/data-sheets/MAX951-MAX954.pdf https://www.analog.com/media/en/technical-documentation/data-sheets/MAX951-MAX954.pdf https://pressbooks.pub/sound/chapter/helmholtz-resonators/ https://pressbooks.pub/sound/chapter/helmholtz-resonators/ Structure Assisted Spectrum Sensing for Low-power Acoustic Event Detection CPS-IoT Week Workshops ’23, May 09–12, 2023, San Antonio, TX, USA imaging via compressive sampling. IEEE signal processing magazine 25, 2 (2008), 83–91. [16] Edward E Fenimore and Thomas M Cannon. 1978. Coded aperture imaging with uniformly redundant arrays. Applied optics 17, 3 (1978), 337–347. [17] Zhihui Gao, Ang Li, Dong Li, Jialin Liu, Jie Xiong, Yu Wang, Bing Li, and Yiran Chen. 2022. MOM: Microphone based 3D Orientation Measurement. In 2022 21st ACM/IEEE International Conference on In- formation Processing in Sensor Networks (IPSN). IEEE, 132–144. [18] Nakul Garg, Yang Bai, and Nirupam Roy. 2021. Demo: Microstructure- guided Spatial Sensing for Low-power IoT. In The 19th Annual Interna- tional Conference on Mobile Systems, Applications, and Services (MobiSys ’21). [19] Nakul Garg, Yang Bai, and Nirupam Roy. 2021. Owlet: Enabling spatial information in ubiquitous acoustic devices. In Proceedings of the 19th Annual International Conference on Mobile Systems, Applications, and Services. 255–268. [20] Nakul Garg and Nirupam Roy. 2020. Enabling self-defense in small drones. In Proceedings of the 21st International Workshop on Mobile Computing Systems and Applications. 15–20. [21] Jason Geng. 2011. Structured-light 3D surface imaging: a tutorial. Advances in Optics and Photonics 3, 2 (2011), 128–160. [22] GrahamMGibson, Steven D Johnson, andMiles J Padgett. 2020. Single- pixel imaging 12 years on: a review. Optics express 28, 19 (2020), 28190– 28208. [23] Uno Ingard. 2009. Noise reduction analysis. Jones & Bartlett Publishers. [24] David Jun, Ondrej Nespesny, Jan Pencik, Zuzana Fisarova, and Ales Rubina. 2021. Optimized method for Helmholtz resonator design formed by perforated boards. Applied Acoustics 184 (2021), 108341. [25] J Klaus, I Bork, M Graf, and G-P Ostermeyer. 2014. On the adjustment of Helmholtz resonators. Applied acoustics 77 (2014), 37–41. [26] HJ Landau. 1967. Sampling, data transmission, and the Nyquist rate. Proc. IEEE 55, 10 (1967), 1701–1706. [27] Anat Levin, Rob Fergus, Frédo Durand, and William T Freeman. 2007. Image and depth from a conventional camera with a coded aperture. ACM transactions on graphics (TOG) 26, 3 (2007), 70–es. [28] Dingzeyu Li, David IW Levin, Wojciech Matusik, and Changxi Zheng. 2016. Acoustic voxels: Computational optimization ofmodular acoustic filters. ACM Transactions on Graphics (TOG) 35, 4 (2016), 1–12. [29] Jialin Liu, Dong Li, Lei Wang, and Jie Xiong. 2021. BlinkListener: " Listen" to Your Eye Blink Using Your Smartphone. Proceedings of the ACM on Interactive, Mobile, Wearable and Ubiquitous Technologies 5, 2 (2021), 1–27. [30] Wenguang Mao, Jian He, and Lili Qiu. 2016. Cat: high-precision acous- tic motion tracking. In Proceedings of the 22nd Annual International Conference on Mobile Computing and Networking. 69–81. [31] Wenguang Mao, Mei Wang, and Lili Qiu. 2018. Aim: Acoustic imaging on a mobile. In Proceedings of the 16th Annual International Conference on Mobile Systems, Applications, and Services. 468–481. [32] Wenguang Mao, Zaiwei Zhang, Lili Qiu, Jian He, Yuchen Cui, and Sangki Yun. 2017. Indoor follow me drone. In Proceedings of the 15th annual international conference on mobile systems, applications, and services. 345–358. [33] James B Mehl. 1978. Analysis of resonance standing-wave measure- ments. The Journal of the Acoustical Society of America 64, 5 (1978), 1523–1525. [34] Kosuke Nagaya, Yasudide Hano, and Akihisa Suda. 2001. Silencer consisting of two-stage Helmholtz resonator with auto-tuning control. the Journal of the Acoustical Society of America 110, 1 (2001), 289–295. [35] Rajalakshmi Nandakumar, Shyamnath Gollakota, and Nathaniel Wat- son. 2015. Contactless sleep apnea detection on smartphones. In Pro- ceedings of the 13th annual international conference on mobile systems, applications, and services. 45–57. [36] Rajalakshmi Nandakumar, Vikram Iyer, Desney Tan, and Shyamnath Gollakota. 2016. Fingerio: Using active sonar for fine-grained finger tracking. In Proceedings of the 2016 CHI Conference on Human Factors in Computing Systems. 1515–1525. [37] Ramesh Raskar, Amit Agrawal, and Jack Tumblin. 2006. Coded expo- sure photography: motion deblurring using fluttered shutter. In Acm Siggraph 2006 Papers. 795–804. [38] Luis Robles and Mario A Ruggero. 2001. Mechanics of the mammalian cochlea. Physiological reviews 81, 3 (2001), 1305–1352. [39] V Romero-García, Noe Jimenez, G Theocharis, V Achilleos, AMerkel, O Richoux, V Tournat, J-P Groby, and V Pagneux. 2020. Design of acoustic metamaterials made of Helmholtz resonators for perfect absorption by using the complex frequency plane. Comptes Rendus. Physique 21, 7-8 (2020), 713–749. [40] A Selamet, FD Denia, and AJ Besa. 2003. Acoustic behavior of circular dual-chamber mufflers. Journal of Sound and Vibration 265, 5 (2003), 967–985. [41] Irtaza Shahid, Yang Bai, Nakul Garg, and NirupamRoy. 2022. VoiceFind: Noise-Resilient Speech Recovery in Commodity Headphones. In Pro- ceedings of the 1st ACM International Workshop on Intelligent Acoustic Systems and Applications. 13–18. [42] Edgar AG Shaw. 1997. Acoustical features of the human external ear. Binaural and spatial hearing in real and virtual environments 25 (1997), 47. [43] Sheng Shen, Daguan Chen, Yu-Lin Wei, Zhijian Yang, and Romit Roy Choudhury. 2020. Voice localization using nearby wall reflections. In Proceedings of the 26th Annual International Conference on Mobile Computing and Networking. 1–14. [44] Xingzhe Song, Boyuan Yang, Ge Yang, Ruirong Chen, Erick Forno, Wei Chen, andWei Gao. 2020. SpiroSonic: monitoring human lung function via acoustic sensing on commodity smartphones. In Proceedings of the 26th Annual International Conference on Mobile Computing and Networking. 1–14. [45] Vamsi Talla, Bryce Kellogg, Shyamnath Gollakota, and Joshua R Smith. 2017. Battery-free cellphone. Proceedings of the ACM on Interactive, Mobile, Wearable and Ubiquitous Technologies 1, 2 (2017), 1–20. [46] PK Tang and WA Sirignano. 1973. Theory of a generalized Helmholtz resonator. Journal of Sound and Vibration 26, 2 (1973), 247–262. [47] Hermann Von Helmholtz. 1912. On the Sensations of Tone as a Physio- logical Basis for the Theory of Music. Longmans Green. [48] Wei Wang, Alex X Liu, and Ke Sun. 2016. Device-free gesture tracking using acoustic signals. In Proceedings of the 22nd Annual International Conference on Mobile Computing and Networking. 82–94. [49] Yangbo Xie, Tsung-Han Tsai, Adam Konneker, Bogdan-Ioan Popa, David J Brady, and Steven ACummer. 2015. Single-sensor multispeaker listening with acoustic metamaterials. Proceedings of the National Academy of Sciences 112, 34 (2015), 10595–10598. [50] Shilong Zhai, Huaijun Chen, Changlin Ding, Fangliang Shen, Chun- rong Luo, and Xiaopeng Zhao. 2015. Manipulation of transmitted wave front using ultrathin planar acoustic metasurfaces. Applied Physics A 120 (2015), 1283–1289. [51] Dan Zhao and XY Li. 2015. A review of acoustic dampers applied to combustion chambers in aerospace industry. Progress in Aerospace Sciences 74 (2015), 114–130. [52] Hongfei Zhu and Fabio Semperlotti. 2013. Metamaterial based em- bedded acoustic filters for structural applications. AIP Advances 3, 9 (2013), 092121. 284 Abstract 1 Introduction 2 Application scenarios 3 Theoretical Foundations 3.1 Passive filtering 3.2 Standing wave resonance 3.3 Helmholtz resonance 4 Lyra Architecture 5 Feasibility Study 6 Limitation and Future Work 7 Related Work 8 Conclusion References