Hz Signal Research Articles

Statistical quantitative assessment of stress in steel strands via MMM technique

The accurate monitoring and assessment of the stress state for steel strands is essential to ensure the safe service of engineering structures. In this study, the metal magnetic memory (MMM) technique was innovatively employed to assess the stress in steel strands. The MMM measurement experiments under tensile loads of the 8 steel strands were conducted. The variation laws of the Hz signals (the normal component of MMM signals) with tensile loads were analyzed from theoretical analysis and experimental results. The quantitative exponential relationship between the evaluation index Iσ and the tensile load F was proposed, and the mean value of the goodness-of-fit was 0.943. Finally, based on the Bayesian method, the stress in steel strands was quantitatively assessed via the normalized index G from the perspective of probabilistic statistics. The mean value of the posterior probability reached 67.9%. This study provides a new method for stress assessment of steel strands.

State-of-the-art sleep arousal detection evaluated on a comprehensive clinical dataset

Aiming to apply automatic arousal detection to support sleep laboratories, we evaluated an optimized, state-of-the-art approach using data from daily work in our university hospital sleep laboratory. Therefore, a machine learning algorithm was trained and evaluated on 3423 polysomnograms of people with various sleep disorders. The model architecture is a U-net that accepts 50 Hz signals as input. We compared this algorithm with models trained on publicly available datasets, and evaluated these models using our clinical dataset, particularly with regard to the effects of different sleep disorders. In an effort to evaluate clinical relevance, we designed a metric based on the error of the predicted arousal index. Our models achieve an area under the precision recall curve (AUPRC) of up to 0.83 and F1 scores of up to 0.81. The model trained on our data showed no age or gender bias and no significant negative effect regarding sleep disorders on model performance compared to healthy sleep. In contrast, models trained on public datasets showed a small to moderate negative effect (calculated using Cohen's d) of sleep disorders on model performance. Therefore, we conclude that state-of-the-art arousal detection on our clinical data is possible with our model architecture. Thus, our results support the general recommendation to use a clinical dataset for training if the model is to be applied to clinical data.

Aerosol interference with open-path eddy covariance measurement in a lake environment

Open-path eddy covariance (OPEC) is widely used for measuring trace gas fluxes between the surface and the atmosphere. At a lake in Eastern China (Lake Taihu), the CO2 flux measured with OPEC was often negative at night (with values as low as −22.1 μmol m−2 s−1) and was coherent across the whole lake, as if the lake were a large sink of atmospheric CO2. The purpose of this study is to investigate the cause of this negative flux phenomenon. In addition to OPEC, we also used closed-path eddy covariance (CPEC) and the transfer coefficient (TC) method to measure the flux. The results show that the persistent negative CO2 flux phenomenon was observed with OPEC but not with CPEC or TC. Because air drawn into the CPEC analyzer was filtered but the OPEC analyzer was influenced by aerosol contamination, the most logical explanation was that particles deposited on the optical lens of the OPEC analyzer changed its cross-sensitivity to water vapor. The direct evidence of this interference was a strong positive correlation between the OPEC analyzer's signal strength and the CO2 mixing ratio observed at 10 Hz. We suggest that it is possible to perform post-field correction to this negative flux bias using the 10 Hz signal strength data. In comparison, an OPEC system at a nearby land site did not experience aerosol interference due to low water vapor flux at night and lack of hygroscopic growth of particles on the optical lens in low humidity conditions. The type of aerosol interference reported here may also occur in high humidity and high pollution conditions elsewhere.

Evaluation of residual load-bearing capacity for corroded steel strands via MMM technique

The quantitative evaluation of the residual load-carrying capacity of corroded cables is critical to the safe service of bridge structures. In this study, the locally corroded steel strands with different corrosion degrees were obtained through electrochemical experiments and the metal magnetic memory (MMM) signals were collected by the TSC-9 M-12 tester. The tensile failure tests of corroded steel strands were conducted by the WAW-1000 tensile testing machine. The distribution laws of the normal component Hz signals for steel strands in different corrosion cases were detailed and a three-dimensional (3D) magnetic dipole model of corroded steel strands was developed considering the end effects. The linear relationships between the magnetic characteristic quantity Hn and the corrosion cross-section loss rate As were revealed and verified by the theoretical and finite element simulation results. Based on the above relationships, a quantitative linear relationship between the residual load-carrying capacity F and the corrosion cross-section loss rate As of the corroded steel strand was proposed. The relative errors of the residual load-carrying capacity F calculated by the proposed quantitative method were overall within 20%. This study provides a basis for evaluating the residual load-carrying capacity of corroded cables using the MMM technique.

A 22.3-Bit Third-Order Delta-Sigma Modulator for EEG Signal Acquisition Systems

This paper presents a high resolution delta-sigma modulator for continuous acquisition of electroencephalography (EEG) signals. The third-order single-loop architecture with a 1-bit quantizer is adopted to achieve 22.3-bit resolution. The effects of thermal noise on the performance of the delta-sigma modulator are analyzed to reasonably allocate the switched-capacitor sizes for optimal signal to noise ratio (SNR) and minimum chip area. The coefficients in feedback path and input path are optimized to avoid the signal distortion under the full-scale input voltage range with almost no increase in total capacitance sizes. Fabricated in 0.5 µm CMOS technology and powered by a 5 V voltage supply, the proposed delta-sigma modulator can achieve 136 dB peak SNR with 16 Hz input and 137 dB dynamic range in 100 Hz signal bandwidth with an oversampling ratio of 512. The modulator dissipates 700 µA. The core chip area is 1.96 mm2. The modulator occupies 1.41 mm2 and the decimator occupies 0.55 mm2.

Poster Session I: Binocular contrast integration: Cortical and behavioral signals reflect different computations.

Although binocular contrast perception under dichoptic viewing conditions has been extensively characterized behaviorally, little is known about how the signals from each eye are combined in cortex. Here we compared simultaneously-collected behavioral and EEG measures of dichoptic contrast perception. Observers (n=16) dichoptically viewed a 2-cpd grating flickering at 7.5 Hz and rotating slowly (1º/s). The contrast of the grating shown to each eye modulated sinusoidally over time at independent rates (1/6 and 1/8 Hz). We recorded EEG activity while observers positioned a joystick lever to report perceived contrast as it changed over time. A multiband filter was used to isolate the SSVEP responses to 7.5 Hz signals from electrode Oz (occipital pole), and the standard deviation of this signal provided a measure of neural response amplitude as a function of contrast. A simple model was fit to VEP and behavioral responses, [(L^m+R^m)/2]^(1/m)], which essentially characterized whether responses were better fit by a mean (m≈1) or a max (m>1) model. VEP responses (m = 1.0; MSE = 0.013) were well fit by a mean model, suggesting the EEG signal may have been driven by the input layers of V1. In contrast, behavioral responses (m = 8.8; MSE = 0.011) were well fit by a model that was heavily shifted towards a max model, suggesting a significant non-linear transformation of the contrast signal between the input layers of V1 and conscious perception.

Detecting the Kauai source beacon with ocean observatories innitiative hydrophones

In March of 2023, a sound source off the coast of Kauai began regularly transmitting a 75 Hz signal into the ocean, six times a day, every fourth day. The regular transmission of this signal allows for the long-term measurements of ocean basin scale temperature changes using the technique of acoustic tomography, as well as further investigations of long-distance acoustic propagation. The Ocean Observatory Initiative (OOI) has, as part of its suite of oceanographic sensors, 11 hydrophones that are continually recording ambient sound. These hydrophones are in various diverse environments in the north-east pacific including bottom mounted in shallow environments, deep environments, on top of seamounts, and moored at 200 m below the ocean surface. In this talk, we present preliminary results of acoustic propagation modeling and data analysis towards receiving the Kauai signal with the OOI hydrophones. The diverse environments of the hydrophones allow us to investigate how different levels of ambient sound and environmental factors, such as proximity to the sound channel affect the ability to detect a positive reception of the Kauai source across the ocean basin. [Work supported by ONR.]

Energy Measurement Technology of Low-frequency 20 Hz Intelligent Energy Meter

A high-speed and high-precision voltage and current acquisition scheme suitable for 20 Hz low-frequency energy metering, and an energy meter suitable for low-frequency transmission scenarios was designed in this paper. A high-precision energy metering algorithm based on low-frequency metering for three-phase energy meters, using the four-term third-order Nuttall window algorithm to address the impact of asynchronous sampling on metering was proposed in this paper. Equipped with sampling and metering algorithms suitable for low-frequency transmission, the designed meter achieves seamless calibration of diagonal difference and stable power calculation. Adopting an A/D+DSP metering scheme, a 24-bit A/D chip is selected as the signal sampling unit, and a 32-bit high-speed DSP is used as the signal processing unit to achieve the sampling and processing function of 20 Hz signals, completing the low-frequency transmission scenario electricity metering.

Repeatability of Pulse Oximetry Measurements in Children During Triage in 2 Ugandan Hospitals.

In low- and middle-income countries, health workers use pulse oximeters for intermittent spot measurements of oxygen saturation (SpO2). However, the accuracy and reliability of pulse oximeters for spot measurements have not been determined. We evaluated the repeatability of spot measurements and the ideal observation time to guide recommendations during spot check measurements. Two 1-minute measurements were taken for the 3,903 subjects enrolled in the study conducted April 2020-January 2022 in Uganda, collecting 1 Hz SpO2 and signal quality index (SQI) data. The repeatability between the 2 measurements was assessed using an intraclass correlation coefficient (ICC), calculated using a median of all seconds of non-zero SpO2 values for each recording (any quality, Q1) and again with a quality filter only using seconds with SQI 90% or higher (good quality, Q2). The ICC was also recalculated for both conditions of Q1 and Q2 using the initial 5 seconds, then the initial 10 seconds, and continuing with 5-second increments up to the full 60 seconds. Lastly, the whole minute ICC was calculated with good quality (Q2), including only records where both measurements had a mean SQI of more than 70% (Q3). The repeatability ICC with condition Q1 was 0.591 (95% confidence interval [CI]=0.570, 0.611). Using only the first 5 seconds of each measurement reduced the repeatability to 0.200 (95% CI=0.169, 0.230). Filtering with Q2, the whole-minute ICC was 0.855 (95% CI=0.847, 0.864). The ICC did not improve beyond the first 35 seconds. For Q3, the repeatability rose to 0.908 (95% CI=0.901, 0.914). Training guidelines must emphasize the importance of signal quality and duration of measurement, targeting a minimum of 35 seconds of adequate-quality, stable data. In addition, the design of new devices should incorporate user prompts and force quality checks to encourage more accurate pulse oximetry measurements.

Characterization of two-dimensional cartridge type bidirectional proportional throttle valve

A novel bidirectional proportional throttle valve with a two-dimensional (2D) plug-in structure is proposed in this paper. The valve integrates the 2D valve module and the logic valve into one component, with high integration degree, response speed, and control accuracy. The mathematical model of the 2D valve module is established based on the continuity equation of flow and kinematic equation, and the stability criterion of the valve is established. The effects of the inclination angle, width, and initial overlapping height of the oblique slot and rectangular slot on the dynamic response of the valve are explored by AMESim simulation, and appropriate structural parameters are selected to design and manufacture the prototype valve. A dedicated test bench is built, and the feasibility of the valve design is verified by testing the valve under different system pressures and different load variations. The experimental results show that, at a system pressure of 21 MPa, the prototype valve has a flow rate of 53.5 L min−1, a hysteresis of 0.8%, a linearity of 5%, and a step response time of 6.5 ms. When the analog valve opening is changed by inputting 0.1 Hz signals with different waveforms under different system pressures, the maximum fluctuation of the main spool displacement is less than 1 μm, and the maximum stabilization time does not exceed 110 ms.

Passive Power Line Communication Filter Design and Benchmarking Using Scattering Parameters

NB-PLC (narrowband power line communication) is a method of data communication that involves superimposing a relatively high-frequency signal (9 kHz to 500 kHz), which contains data, onto the power grid’s low frequency (50 to 60 Hz) signal. While using the existing power grid as a transmission medium is convenient, the power grid was not designed for this purpose, leading to challenges such as conducted emissions and infrastructure limitations. To overcome these technical challenges, passive filters are necessary. This article presents the design, simulation (using scattering parameters), and evaluation of an NB-PLC filter by comparing it to commercially available filters. Our proposed design and benchmarking methods enable the accurate prediction of the filter’s behavior in field conditions. After comparing our filter with commercially available filters, we observed that it exhibits superior characteristics. Specifically, our filter has the best insertion loss versus frequency, achieved three times higher attenuation at 50 kHz (−130 dB) compared to the best commercially available filter (−40 dB), and has a power consumption of 0.6 W, which is comparable to the most power-efficient commercial filter (0.5 W). Additionally, our filter has the second best input and output impedance of 3.6 Ω within the frequency range of 35–95 kHz.

Simulation of a new respiratory phase sorting method for 4D-imaging using optical surface information towards precision radiotherapy

BackgroundRespiratory signal detection is critical for 4-dimensional (4D) imaging. This study proposes and evaluates a novel phase sorting method using optical surface imaging (OSI), aiming to improve the precision of radiotherapy. MethodBased on 4D Extended Cardiac-Torso (XCAT) digital phantom, OSI in point cloud format was generated from the body segmentation, and image projections were simulated using the geometries of Varian 4D kV cone-beam-CT (CBCT). Respiratory signals were extracted respectively from the segmented diaphragm image (reference method) and OSI respectively, where Gaussian Mixture Model and Principal Component Analysis (PCA) were used for image registration and dimension reduction respectively. Breathing frequencies were compared using Fast-Fourier-Transform. Consistency of 4DCBCT images reconstructed using Maximum Likelihood Expectation Maximization algorithm was also evaluated quantitatively, where high consistency can be suggested by lower Root-Mean-Square-Error (RMSE), Structural-Similarity-Index (SSIM) value closer to 1, and larger Peak-Signal-To-Noise-Ratio (PSNR) respectively. ResultsHigh consistency of breathing frequencies was observed between the diaphragm-based (0.232 Hz) and OSI-based (0.251 Hz) signals, with a slight discrepancy of 0.019Hz. Using end of expiration (EOE) and end of inspiration (EOI) phases as examples, the mean±1SD values of the 80 transverse, 100 coronal and 120 sagittal planes were 0.967, 0,972, 0.974 (SSIM); 1.657 ± 0.368, 1.464 ± 0.104, 1.479 ± 0.297 (RMSE); and 40.501 ± 1.737, 41.532 ± 1.464, 41.553 ± 1.910 (PSNR) for the EOE; and 0.969, 0.973, 0.973 (SSIM); 1.686 ± 0.278, 1.422 ± 0.089, 1.489 ± 0.238 (RMSE); and 40.535 ± 1.539, 41.605 ± 0.534, 41.401 ± 1.496 (PSNR) for EOI respectively. ConclusionsThis work proposed and evaluated a novel respiratory phase sorting approach for 4D imaging using optical surface signals, which can potentially be applied to precision radiotherapy. Its potential advantages were non-ionizing, non-invasive, non-contact, and more compatible with various anatomic regions and treatment/imaging systems.

A Wirelessly-Powered Body-Coupled Data Transmission with Multi-Stage and Multi-Source Rectifier

This paper demonstrates body-coupled (BC) data transmission and multi-source power delivery systems for neural interface applications. The implanted data transmitter and power receiver utilize an electrode interface rather than an antenna or coil interface for battery-free wireless transmission, enabling the external data receiver and power transmitter with patch electrodes to be placed away from the implant without requiring precise alignment, which is a critical issue in the conventional communication modalities of inductive coupling. Significantly, the implanted power receiver produces the supply voltage using ambient body-coupled 50/60 Hz signals from the Mains, on top of the 40.68 MHz wireless power source from the external power transmitter, to increase the recovered power level and the voltage conversion efficiency (VCE). The body-coupled wireless systems for implanted and external devices are implemented with integrated circuits (ICs) fabricated in a 180 nm CMOS process. When 650 mVpp AC voltage is applied to the implanted device, the power receiver recovers up to 780 μW with ambient (60 Hz signals) BC energy harvesting, achieving 93% VCE, while 600 μW is recovered without ambient (60 Hz signals) signal. The recovered power supplies the regulated voltage to the direct-digital signaling transceiver, which consumes 460 μW with an uplink data rate of 10 Mbps and a downlink data rate of 200 kbps, corresponding to an energy efficiency of 46 pJ/b.

A presentation attack detection network based on dynamic convolution and multi-level feature fusion with security and reliability

In the Internet of Things environment, most facial recognition systems are highly susceptible to face spoofing attacks. To increase the security and reliability of such systems, this paper proposes an anti-spoofing method for facial recognition systems based on optical flow and texture features. The spoofing-detection algorithm first generates the optical flow field map of the face area using the optical flow method and a face-detection method, based on two consecutive frames of the captured face video. Then, the original RGB face area image and optical flow field map are input into a two-channel convolutional neural network to extract and fuse the features of the face. Finally, based on the optical flow and texture features, this method classifies real and fake faces. In addition, to better generate the optical flow field map containing liveness information, a motion amplification algorithm is applied to enhance the 0.04–0.4 Hz signal in the video frame by 20 times. For texture and optical flow presentation, we propose a lightweight network with dynamic convolution and multi-level feature fusion blocks. This paper used the Replay Attack spoofing dataset from IDIAP, consisting of 1300 videos, for model training, verification, and testing. Experiments showed that the algorithm proposed in the paper performed well on the Replay Attack data set and achieved a half total error rate of 0.66%.

A Novel Method for ECG Signal Compression and Reconstruction: Down-Sampling Operation and Signal-Referenced Network

Objective. The wearable electrocardiogram (ECG) monitoring device is an effective tool for diagnosing intermittent heart diseases. However, the massive amount of ECG data increases power consumption during wireless transmission, thereby reducing the monitoring duration of the wearable device. Here, we presented a novel method to solve this problem. Method. To reduce the data size, we used a down-sampling operation to compress signals. Then, we designed a signal-referenced network to reconstruct the original signals from the compressed ones. We validated the proposed method on the China Physiological Signal Challenge 2018 database, used the root mean square error (RMSE) to evaluate the performance of the proposed network, and evaluated the effectiveness of the reconstructed signals via the F1-score of an ECG signal classifier. Main Result. The classifier used in this paper achieved an F1-score of 84% on 500 Hz signals reconstructed from 25 Hz, 89% from 50 Hz, 90% from 125 Hz, and 95% from 250 Hz. The RMSE of these four sampling rates was 0.10 mV, 0.08 mV, 0.05 mV, and 0.04 mV, respectively. Significance. The experimental result shows that the proposed network has a good performance when reconstructing signals. Furthermore, our method can remove the computational load of compressing signals from wearable devices.

448 Analysis of Bipolar Re-Referencing on Spectral Composition and Spike Detection in Human Electrocorticography

INTRODUCTION: Electrocorticography (ECoG) enables the localization of epileptic seizure foci prior to surgical resection. Improving the signal-to-noise ratio is paramount, and can be accomplished with bipolar re-referencing. METHODS: Intracranial recordings for drug-resistant epilepsy were performed at UCSF using high-density ECoG grid (4 mm inter-electrode pitch), strip (10 mm), and penetrating depth (5 mm) electrodes. Raw voltages for each participant (n = 11) were systematically re-referenced (bipolar subtraction) at increasing scales (4 mm-50 mm) and a multi-taper power spectrum (2-200 Hz) was generated around spike-free baseline windows. Cluster based permutation tests (p < 0.05) were used to compare spectral composition features after z-scoring log power features by frequency and binning by bipolar distance. High-density and low-density (4mm vs 8mm) re-referencing schemes were used for spike localization by applying a line length transform to highlight epileptiform features in peri-spike windows and comparing to baseline windows using cluster-based permutation tests. RESULTS: Bipolar pair distances <10 mm decreased low-frequency (<30 Hz) power, whereas both low and high frequency (30-200 Hz) power increased by longer bipolar distances. These findings were consistent across different components (grids, strips, depths). The detection of high gamma (50-200 Hz) neural activity in the STG related to an audio stimulus was enhanced at bipolar distances of <45 mm but attenuated at higher distances. Similar spike detection (duration and number of channels involved) was seen between high- vs. low-density bipolar re-referencing. CONCLUSIONS: Re-referencing at <10 mm bipolar distance preferentially attenuated low-frequency signal while standard clinical and larger distances systematically augmented broadband (2-200 Hz) signals. Shorter distances, especially high-density, augmented spatial delineation of neural activity when compared to more widely spaced arrays, however, standard clinical arrays capture the general spatio-temporal extent of spike waveforms.

Accretion-induced Collapse of Dark Matter-admixed Rotating White Dwarfs: Dynamics and Gravitational-wave Signals

We present two-dimensional hydrodynamic simulations of the accretion-induced collapse (AIC) of rotating white dwarfs admixed with an extended component of dark matter (DM) comprising sub-gigaelectronvolt degenerate fermionic DM particles. We find that the DM component follows the collapse of the normal matter (NM) component to become a bound DM core. Thus, we demonstrate how a DM-admixed neutron star could form through DM-admixed AIC (DMAIC) for the first time, with the dynamics of DM taken into account. The gravitational-wave (GW) signature from the DMAIC shows distinctive features. In the diffusive DM limit, the DM admixture indirectly suppresses the post-bounce spectral peak of the NM GWs. In the compact DM limit, the collapse dynamics of the DM in a Milky Way event generate GWs that are strong enough to be detectable by Advanced LIGO as continuous low-frequency (<1000 Hz) signals after the NM core bounce. Our study not only is the first-ever computation of GW from a collapsing DM object but also provides the key features to identify DM in AIC events through future GW detections.

The sound of the ingenuity Helicopter on Mars

The sounds of the Ingenuity Helicopter flying in the Martian atmosphere are among the most notable recordings of the microphone on the SuperCam instrument on the Mars 2020 Perseverance Rover. Distinct acoustic signatures of the helicopter were recorded on the 4th, 5th, 6th, and 8th flights. The detected signatures are around 84 Hz and (occasionally) at 168 Hz, at the blade crossing frequency and its first harmonic. Several higher harmonics were prominent in hover tests in short-range recordings in a test chamber on Earth; these are attenuated by CO2 absorption at the 50 m-plus ranges on Mars. Doppler shift of the 84 Hz signal can be measured and is consistent with the trajectory measured with Ingenuity’s navigation camera and inertial navigation unit, and documented by Perseverance’s cameras. A striking feature of the sound recordings is an unanticipated deep modulation of the signals with nulls spaced by around 15–20 s, superposed on the simple and expected decline in amplitude with distance. We have evaluated and rejected models of multipath sound interference as requiring implausibly strong near-surface temperature gradients. We find instead that the modulation appears to be the signature of a slight asynchrony between the rotation rates of the two coaxial rotors, such that the blade-crossing azimuth rotates slowly during flight, resulting in a “lighthouse” sweeping of the radiated sound pattern. Analysis of blade orientations seen in the shadow of the helicopter observed in down-looking navigation images supports this model.

Identifying the distinct spectral dynamics of laminar-specific interhemispheric connectivity with bilateral line-scanning fMRI

Despite extensive efforts to identify interhemispheric functional connectivity (FC) with resting-state (rs-) fMRI, correlated low-frequency rs-fMRI signal fluctuation across homotopic cortices originates from multiple sources. It remains challenging to differentiate circuit-specific FC from global regulation. Here, we developed a bilateral line-scanning fMRI method to detect laminar-specific rs-fMRI signals from homologous forepaw somatosensory cortices with high spatial and temporal resolution in rat brains. Based on spectral coherence analysis, two distinct bilateral fluctuation spectral features were identified: ultra-slow fluctuation (<0.04 Hz) across all cortical laminae versus Layer (L) 2/3-specific evoked BOLD at 0.05 Hz based on 4 s on/16 s off block design and resting-state fluctuations at 0.08–0.1 Hz. Based on the measurements of evoked BOLD signal at corpus callosum (CC), this L2/3-specific 0.05 Hz signal is likely associated with neuronal circuit-specific activity driven by the callosal projection, which dampened ultra-slow oscillation less than 0.04 Hz. Also, the rs-fMRI power variability clustering analysis showed that the appearance of L2/3-specific 0.08–0.1 Hz signal fluctuation is independent of the ultra-slow oscillation across different trials. Thus, distinct laminar-specific bilateral FC patterns at different frequency ranges can be identified by the bilateral line-scanning fMRI method.

Efficient Development of Gait Classification Models for Five-Gaited Horses Based on Mobile Phone Sensors.

Automated gait classification has traditionally been studied using horse-mounted sensors. However, smartphone-based sensors are more accessible, but the performance of gait classification models using data from such sensors has not been widely known or accessible. In this study, we performed horse gait classification using deep learning models and data from mobile phone sensors located in the rider's pocket. We gathered data from 17 horses and 14 riders. The data were gathered simultaneously from movement sensors in a mobile phone located in the rider's pocket and a gait classification system based on four wearable sensors attached to the horse's limbs. With this efficient approach to acquire labelled data, we trained a Bi-LSTM model for gait classification. The only input to the model was a 50 Hz signal from the phone's accelerometer and gyroscope that was rotated to the horse's frame of reference. We demonstrate that sensor data from mobile phones can be used to classify the five gaits of the Icelandic horse with up to 94.4% accuracy. The result suggests that horse riding activities can be studied at a large scale using mobile phones to gather data on gaits. While our study showed that mobile phone sensors could be effective for gait classification, there are still some limitations that need to be addressed in future research. For example, further studies could explore the effects of different riding styles or equipment on gait classification accuracy or investigate ways to minimize the influence of factors such as phone placement. By addressing these questions, we can continue to improve our understanding of horse gait and its role in horse riding activities.