Traditional SONAR systems are widely used for underwater object detection and navigation; however, they suffer from high energy consumption, noise interference, and signal degradation in varying aquatic conditions. Inspired by biological echolocation, we propose a novel machine learning-enhanced sonar model to improve accuracy, robustness, and energy efficiency in underwater sensing applications. The proposed model employs bio-inspired echolocation principles, where transmitted acoustic pulses dynamically adjust in response to environmental conditions. Deep learning techniques, specifically Convolutional Neural Networks (CNNs) and Long Short-Term Memory (LSTM) networks, are employed to classify sonar echoes and enhance object detection. Additionally, digital signal processing (DSP) techniques, including Butterworth filtering, wavelet decomposition, and adaptive thresholding, are integrated to mitigate noise and improve signal clarity. Experimental evaluations demonstrate that the proposed sonar model, SonarNet, achieves 92.7% classification accuracy, outperforming conventional SONAR-based methods, which achieved 85.4% accuracy. The integration of adaptive signal processing leads to a 20.8% reduction in energy consumption compared to standard SONAR models, improves the Signal-to-Noise Ratio (SNR) by 15.3dB, and reduces the false positive rate by 18.6%. Additionally, the sound velocity profile (SVP) correction mechanism improves depth estimation accuracy by 12.4%. All performance results reported in this work were obtained using a simulation environment parameterized by real-world oceanographic datasets.

Aging is associated with structural and functional changes of the vocal folds that may result in presbyphonia, often perceived as a weak or shaky voice. However, the quantitative characterization of underlying age-related vocal tremor across the adult lifespan remains limited. This cross-sectional study investigated the characteristics of vocal tremor across the adult lifespan using automated acoustic analysis. A total of 291 native speakers aged 18-94 years were recruited and underwent perceptual voice evaluation and acoustic analysis during sustained phonation of the vowel /a/. Vocal tremor was quantified using digital signal processing, focusing on the prominence of fundamental frequency tremor (PF0T) and the prominence of amplitude tremor (PAT). A moderate-to-strong positive correlation between age and PF0T was observed in both males and females, indicating increasing instability of fundamental frequency with advancing age. In contrast, PAT did not show a significant age-related increase after correction for multiple comparisons. Perceptual ratings of tremor demonstrated only weak correlations with age but were moderately associated with acoustic measures of tremor. Normative models revealed that physiological tremor in healthy aging remains well below pathological thresholds reported in neurological disorders. These findings indicate that age-related vocal tremor is characterized predominantly by increasing instability of fundamental frequency rather than amplitude modulation, localizing the dominant age effect to laryngeal control of vocal fold tension rather than to respiratory drive. Automated acoustic analysis provides a sensitive and objective method for detecting subtle age-related vocal changes and may support future biomarker development for distinguishing physiological from pathological vocal tremor.

Non-orthogonal multiple access (NOMA) combined with digital subcarrier multiplexing (DSCM) enables dense and flexible coherent passive optical networks (PONs). However, NOMA signal superposition biases pilot-aided carrier-phase recovery (CPR), while bandwidth-limited transceivers impose penalties on edge subcarriers. In this work, we propose a unified digital signal processing (DSP) framework that incorporates dual-pilot CPR with pair-wise maximum-ratio combining (MRC) across adjacent center-edge subcarrier pairs. By inserting pilots into both constituent signals of the superposed NOMA stream, the receiver builds a composite pilot reference that suppresses interference-induced phase bias. The resulting reliability estimate is then used to weight subcarrier branches in MRC, turning quality imbalance into diversity gain. In an experimental demonstration with an 80-Gbaud NOMA-DSCM coherent PON, the dual-pilot CPR achieves more robust phase recovery than conventional pilot-based and blind CPR baselines under NOMA-induced interference. With MRC enabled, the framework further delivers up to 2.29 dB SINR gain over benchmarks at low optical power, reducing average BER by 75.89% for ONU 1 and 30.33% for ONU 2.

Fully integrated photonic continuous-variable quantum key distribution is a promising route toward compact and scalable secure optical links, yet the achievable transmission distance has been limited by the performance of integrated receiver chips and excess noise suppression methods. Here, we demonstrate a Gaussian-modulated continuous-variable quantum key distribution system with an integrated silicon photonic receiver, operating over 60 km. Enabled by a high-clearance, broadband silicon photonic receiver, in conjunction with a robust digital signal processing framework that utilizes a time-domain superposition algorithm and secure dynamic single-tap equalization, the system effectively achieves an asymptotic secret key rate of 1.68 Mbps, and a finite-size secret key rate of 0.80 Mbps at a data block length of 1.55×109. This work extends chip-based continuous-variable quantum key distribution to metropolitan-scale distances, confirming the viability of integrated receivers for large-scale quantum networks.

Large-scale artificial intelligence training demands ultralow-latency, energy-efficient interconnects for massive graphics processing unit clusters. In intensity-modulation/direct-detection links, digital signal processing (DSP) equalization is limited by nonideal equalization caused by phase loss as well as tight power and latency budgets. We present an integrated, programmable optical signal processor (OSP) that functions as a nonlinear universal equalizer and performs all-optical, DSP-free, real-time equalization. A deep reservoir with all-optical readout enables a Vernier scheme with ~1-picosecond (ps) sampling resolution and a tunable memory window. The OSP simultaneously equalizes eight wavelength-division-multiplexing (WDM) channels, delivering 1.6-terabits/second aggregate throughput with <60-picoseconds latency and tens of femtojoules/bit energy consumption. Operating before detection, it provides superior chromatic dispersion compensation, mitigates transceiver bandwidth limits and fiber nonlinearity, and expands the usable WDM window by a factor of 6.8.

The upcoming 6G networks require digital signal processing (DSP) systems which need to adapt their operations for unanticipated changes in communication channel conditions and system environmental factors because the demand for high-speed communication demands both quick response times and dependable service. The fixed processing chains of traditional DSP systems which engineers built for 5G networks together with their requirement for specific hardware usage create limits that prevent effective operation under changing environmental conditions. The paper presents ADaPT-6 (Adaptive DSP for Progressive Transceivers in 6G) as a solution to these obstacles through its entire software-based AI-powered digital signal processing system which enables real-time transceiver operation changes through intelligent learning. The ADaPT-6 system operates through two fundamental components which use FlexiTune Modulation Adaptation to enable systems to select their best modulation methods based on current channel conditions while the Signal State Evolution Engine predicts system behavior to modify system operations of filtering and equalization and synchronization. The framework achieves better spectral efficiency through its implementation of adaptive modulation together with its ability to predict signal states which improves system reliability and operational efficiency. The research shows through extensive testing in practical 6G fronthauling situations that both traditional fixed DSP systems and DSP systems with limited adaptation capabilities face performance difficulties across multiple performance metrics including bit error rate and statistical reliability and throughput and energy efficiency and latency and outage probability. The evaluation process used SNR values ranging from 0 to 20dB while measuring EVM in percentage terms and ACLR in decibels and average throughput in Mbps and latency in milliseconds and energy efficiency in bits per Joule. Quantitative evaluation demonstrates that ADaPT-6 achieves significant performance improvements over conventional DSP frameworks, including a ~ 20.8% increase in throughput (57.37Mbps vs. 47.47 Mbps), ~ 23.7% reduction in latency (20.52ms vs. 26.88ms), and ~ 44.5% improvement in energy efficiency (0.678 vs. 0.469 bits/Joule). Additionally, the framework consistently achieves lower BER and EVM across the entire SNR range (0-20dB), confirming its robustness under dynamic channel conditions. The results confirm that ADaPT-6 operates as an effective digital signal processing solution which works on any hardware platform and can be implemented at any scale. The system functions as the best solution for future AI-native 6G transceiver systems.

In order to realize receiver (Rx) photonic integrated circuits (PICs) applicable to 400 Gb/s-class coherent passive optical network (PON) systems, we integrated tunable wavelength f ilters (TWFs), PIN photodiodes (PIN-PDs), and optical 90 degree hybrids (90HBs) into Rx PICs fabricated on a commercial standard silicon photonics (SiPh) platform. As a result of device evaluation, we observed characteristics indicating the potential for application in 400 Gb/s-class coherent PON systems. Subse quently, we prototyped Rx assemblies in which the Rx PIC and a digital signal processing (DSP) chip are mounted. When the Rx assembly received a 60-Gbaud dual-polarization (DP) quadrature phase shift keying (QPSK) 200 Gb/s C-band optical signal after 2 km loopback transmission, we achieved an optical received power of −24 dBm at a bit error ratio (BER) of 2×10−2, which corresponds to the forward error correction (FEC) limit. Based on this result, the Rx assembly is suitable for a PON system with a maximum transmission distance of 20 km and up to 16 terminations. Similarly, when the Rx assembly received a DP 16 quadrature amplitude modulation (QAM) 400 Gb/s optical signal, we achieved an optical received power of −6 dBm at the FEC limit. Although some issues remain, these results suggest the Rx assembly can be applied to 400 Gb/s-class coherent PON systems. In the future, we will improve the Rx characteristics by reducing insertion loss of the Rx PICs and will prototype and evaluate transmitter (Tx) assemblies.

The next-generation of cryogenic neutrinoless double-beta decay experiments require increasingly fast readout in order to improve background discrimination. These experiments, operated as cryogenic calorimeters at ∼ 10 mK, are usually read out by high-impedance neutron transmutation doped (NTD) thermistors, which provide good energy resolution, but are limited by ∼ 1 ms response times. Superconducting detectors, such as transition-edge sensors (TESs) with a time resolution of ∼ 100 μs, offer superior timing performance over NTD semiconductor bolometers. To make this technology viable for an application to a thousand or more channels, multiplexed readout is necessary in order to minimize the thermal load and radioactive contamination induced by the readout. Frequency-domain multiplexing readout (fMUX) for TESs, previously developed at Berkeley Lab and McGill University, is currently in use for mm-wave telescopes with detector sampling rates in the order of 100 Hz. We demonstrate a new readout system, based on the McGill/Berkeley digital fMux readout, to satisfy the higher bandwidth and noise requirements of the next generation of TES-instrumented cryogenic calorimeters. Each multiplexing readout module comprises 10 superconducting resonators in the 1–5 MHz range and a DC superconducting quantum interference device (DC-SQUID), interfaced to high-speed field programmable gate array (FPGA)-based electronics for digital signal processing and low-latency SQUID feedback. The new readout samples detectors at 156 kHz, three orders of magnitude faster than its cosmology-oriented predecessor, and demonstrates a stable feedback bandwidth of 3 kHz in a real TES-based system.

An extensive review on ground penetrating radar signal processing for pavement evaluation: Advances, AI integration and applications

The article analyzes contemporary approaches to the automated detection of artillery shots in complex acoustic environments and substantiates the feasibility of applying digital signal processing and deep learning methods in acoustic reconnaissance systems. The study examines the physical features of impulse acoustic signal formation during a shot, the influence of background noise, terrain relief, and meteorological factors on sound wave propagation parameters, as well as the challenges of isolating useful events under conditions of intensive interference. Particular attention is paid to the analysis of temporal, energy, and spectral signal features, the formation of spectrogram representations, and their use as an informative basis for convolutional neural networks. The article describes the implementation of a software prototype for an automated detection system that combines an algorithmic audio data processing module with an interactive operator interface. A combined detection index for impulse events is proposed, along with an energy-based false alarm filtering mechanism and a preliminary event classification approach based on an integral confidence assessment. The possibilities of applying transfer learning to adapt pre-trained convolutional neural networks to the task of artillery shot identification are also analyzed. The scientific significance of the study lies in the comprehensive integration of classical sound-ranging methods with modern artificial intelligence algorithms and in the identification of informative feature representations of acoustic signals to improve the accuracy of automatic recognition. The practical value is determined by the possibility of integrating the proposed solutions into counter-battery warfare systems, acoustic monitoring systems, and automated combat control platforms. The obtained results provide a foundation for further research in the development of distributed sensor networks, enhancement of algorithm robustness to interference, and expansion of the functional capabilities of acoustic reconnaissance systems in accordance with the current requirements of the Armed Forces of Ukraine.

Real-time binaural beat synthesis in dynamic acoustic environments is challenged by carrier non-stationarity, interaural phase discontinuities, and processing delay in conventional digital signal processing pipelines. This study proposes a predictive dual-stage neural framework for phase-coherent auditory synthesis under non-stationary acoustic conditions. The framework decouples real-time carrier estimation from phase-coherent signal generation through two specialized modules. An intelligent acoustic sensing module (AI-1) estimates time-varying carrier information across harmonic, fluctuating, and broadband acoustic profiles using a causal neural front-end with an adaptive confidence-driven strategy. A predictive phase-coherent generator (AI-2) then forecasts short-horizon carrier trajectories and drives a discrete-time phase accumulator to maintain continuous phase evolution during binaural beat embedding. Objective evaluation under multiple acoustic profiles and noise conditions shows that the proposed framework maintains strong phase continuity, with a Phase Coherence Factor greater than 0.91, and low artifact levels, with a Signal-to-Artifact Ratio greater than 39.8 dB, under the evaluated conditions. Additional comparisons with conventional DSP baselines, stronger classical F0 estimators, a lightweight neural F0 tracker, and component-wise ablation variants further demonstrate that the performance improvement arises from the combination of adaptive carrier estimation and predictive phase-coherent actuation, rather than from carrier estimation alone. Hardware profiling shows a combined INT8 inference time of 2.4 ms per frame on a resource-constrained Raspberry Pi Zero 2W-class edge device. Importantly, this inference time and the sub-millisecond phase-accumulator resolution should not be interpreted as sub-millisecond end-to-end physical audio latency. The complete system still includes buffering, framing, neural inference, and output processing delay; the proposed method instead reduces effective phase-boundary misalignment through short-horizon predictive compensation. These results support the proposed framework as a lightweight engineering solution for real-time phase-continuous auditory synthesis in dynamic listening environments. The reported PCF and SAR values should be interpreted as signal-level indicators of phase continuity and artifact suppression, rather than as evidence of listener comfort, perceptual preference, or neurophysiological efficacy.

Spirometry is a crucial diagnostic test for pulmonary diseases, where measures including Forced Expiratory Volume in one second (FEV1) have diagnostic and prognostic value. Symptomatic sounds with digital signal processing methods and machine learning techniques have potential for monitoring respiratory disease. Previously, a threshold-based speech and breath extraction method was used to develop a multi-model approach to predict FEV1% predicted from voice recordings. This study further optimises this method to improve predictive performances of models by data augmentation using data segmentation, handling class imbalance, and enhancing predictive models of Random Forest (RF), Logistic Regression (LR) and Support Vector Machine (SVM) with feature selection and hyperparameter tuning. Using 10-second segments with top ten features demonstrates the best results in all three models - regression (RMSE = 9.85), multiclass classification model (Accuracy = 79.28%), and binary classification model with hyperparameter tuning (AUC = 93.57%). The results demonstrate improved performance for the three models after feature selection and hyperparameter tuning. This study presents the limitations of the current experiments and highlights the future work in the ongoing research.

To meet the demand for high-capacity indoor wireless access in future 6G systems, we propose and experimentally demonstrate a photonics-aided D-band wireless transmission scheme operating at 138 GHz. At the transmitter, two external-cavity lasers together with an I/Q modulator are used to generate a modulated D-band carrier. At the receiver, homodyne down-conversion is employed to directly recover the received signal to baseband, thereby relaxing the requirements on ultra-wideband analog components and high-speed sampling hardware. A 20 m indoor line-of-sight wireless link is established to transmit a 56-Gbaud-rate OFDM-QPSK signal. The transmitted and received spectra, received constellations and bit-error-rate (BER) performance are functions of optical power at different symbol rates, and the channel amplitude and phase responses are systematically analyzed. The results show that broadband D-band signal generation, transmission, and recovery can be stably achieved in the proposed system. After receiver-side digital signal processing (DSP), clear QPSK constellations are obtained. BER measurements reveal an optimal optical-power operating range, and the 32-GBaud OFDM signal outperforms the 56-Gbaud-rate signal because its narrower occupied bandwidth makes it less sensitive to frequency-selective distortion. For 56-Gbaud-rate OFDM transmission, the BER approaches the 20% low-density parity-check forward-error-correction threshold at an optical power of approximately −1 dBm. Further analysis indicates that the current link performance is mainly limited by frequency-selective amplitude and phase distortions under bandwidth-constrained conditions, together with slight nonlinear effects at high power. These results verify the feasibility of a photonics-aided D-band wireless architecture with homodyne reception for medium-range, high-symbol-rate indoor transmission and provide an experimental basis for future 6G sub-THz wireless links.

Multipath interference (MPI) severely degrades the performance of high-speed multi-level intensity-modulation/direct-detection (IM/DD) systems by inducing amplitude fluctuations in the received signal, and many digital signal processing (DSP)-based mitigation techniques with high computational complexity have been reported. In this paper, we propose a low-complexity MPI mitigation method based on first-order recursive filtering. The method consists of a coarse mitigation scheme to suppress slowly varying amplitude fluctuations dominated by carrier-carrier beating, and a fine mitigation scheme to both suppress these fluctuations and further reduce residual distortion arising from broadband crosstalk and imperfect coarse-stage suppression. Simulation and experimental results for 32 GBaud PAM-4 systems demonstrate a substantial reduction in bit-error rate (BER) over a wide range of signal-to-interference ratios (SIRs) and laser linewidths. BERs below the KP4-FEC threshold at an SIR of 22 dB are achieved, and the SIR tolerance is improved by over 8 dB using the proposed MPI mitigation method for both broad- and narrow-linewidth laser sources. The proposed scheme provides superior performance compared to previously reported digital filtering methods, while requiring only four multiplications, two additions, and two registers per symbol, thereby yielding ultra-low computational complexity for MPI mitigation.

The textile dyeing sector contributes significantly to excessive water and energy use and environmental pollution. Managing yarn tension during the rewinding process is a crucial challenge in textile dyeing, as it has a direct impact on yarn quality, resource consumption, and overall efficiency. Traditional tension control techniques ignore how tension variation affects the production of wastewater and the need for chemical treatment. Using a digital signal processor (TMS320F28335), this study introduces a novel event-driven yarn tension control method for soft winding machines that maintains ideal, constant tension across fine, medium, and thick cotton yarn counts. The suggested method dramatically reduces tension variance, as demonstrated by experimental validation conducted at a leading Egyptian cotton yarn producer. This decreases yarn defects by 35% and allows for a 27% increase in machine speed without sacrificing yarn quality. Most significantly, liquor ratio data show that optimized tension management reduces water consumption during dyeing by 65–70%, thereby lowering chemical use and wastewater generation. Additionally, energy use dropped by almost 20%, improving process sustainability. This integrated strategy promotes the triple bottom line of social, economic, and environmental advantages and is consistent with cleaner manufacturing principles. The results provide a scalable methodology for environmentally friendly textile production, especially appropriate for developing nations with limited resources.

Background: Sensorineural hearing loss represents a significant global health burden affecting over 1.5 billion individuals worldwide. Modern hearing aids, equipped with digital signal processing and smart connectivity features, constitute a cornerstone of neuro-sensory rehabilitation. However, the psychosocial impact of these assistive smart technologies on patient self-esteem remains incompletely characterized. Methods: A cross-sectional multivariate study was conducted with 245 participants, divided into three groups: normal-hearing controls (NH, n = 73), hearing-impaired patients using smart hearing aid technology (HA users, n = 86), and hearing-impaired patients not using hearing aid technology (HA non-users, n = 86). Self-esteem was measured using the Rosenberg Self-Esteem Scale (SES). Hearing disability and tinnitus severity were assessed with the Hearing Handicap Inventory for Adults (HHIA) and Tinnitus Handicap Inventory (THI), respectively. Data analysis included one-way ANOVA, Tukey’s HSD post hoc tests, Pearson correlations, and multivariate regression. Results: Hearing aid users showed significantly higher SES scores (35.41 ± 5.32) compared to non-users (22.99 ± 4.53; p < 0.001, Cohen’s d = 2.515). One-way ANOVA indicated highly significant differences among groups (F = 299.00, p < 0.001, η2 = 0.712). SES was negatively correlated with HHIA (r = −0.573, p < 0.001) and THI (r = −0.443, p < 0.001), while HHIA and THI were strongly positively correlated (r = 0.729, p < 0.001). In multivariate analysis, HA use remained a strong independent predictor of self-esteem (β ≈ 11.9, p < 0.001), even after adjustment for age, sex, HHIA, and THI. Perceived hearing handicap was independently associated with lower self-esteem, whereas tinnitus severity was not a significant predictor in the fully adjusted model. The model explained approximately 65% of the variance in self-esteem scores. Conclusions: Smart hearing-aid use is strongly and independently associated with higher self-esteem in patients with sensorineural hearing loss. These results support the inclusion of modern audiological rehabilitation devices in comprehensive management strategies for long-term conditions and highlight psychosocial benefits that extend beyond hearing restoration.

Recent advancements in photovoltaic water pumping systems (PVWPS) have garnered significant attention from researchers, driven by their reliance on clean solar energy and their potential for sustainable water management. A typical PVWPS configuration includes a three-phase asynchronous motor (ASM) coupled to a centrifugal pump and powered by a photovoltaic (PV) generator. To optimize power extraction from the PV source, a maximum power point tracking (MPPT) algorithm is employed in conjunction with a boost converter. The system integrates two main control strategies: an MPPT controller for the three-phase inverter and a Field-Oriented Control (FOC) scheme for the asynchronous motor. This study introduces an enhanced perturbation and observation (P&O) MPPT algorithm featuring adaptive step-size control. The proposed method effectively reduces steady-state oscillations and improves the efficiency of power extraction in PV systems. The second control unit, based on FOC, regulates the induction motor by generating the switching signals for the voltage source inverter (VSI). Together, these control units play a vital role in system operation, contributing to improved efficiency and overall performance. The key innovation of this work lies in the integration of the improved P&O MPPT method. Extensive simulation analyses conducted under various irradiation conditions demonstrate that the enhanced algorithm achieves superior tracking performance compared to conventional P&O methods. Furthermore, the effectiveness of the FOC strategy implemented with a two-level inverter has been experimentally validated. The control approach was executed using a dSPACE DS1104 digital signal processor board in an ASM drive system.

Autonomous driving perception demands low latency, high temporal resolution, and stringent hardware efficiency. While event-based spiking neural networks (SNNs) offer bio-inspired sparse computation, their deployment on edge field-programmable gate arrays (FPGAs) is obstructed by irregular execution patterns and temporal state storage overhead. To address this, we propose HAPQ, a unified hardware-aware pruning and quantization pipeline for compact event-based object detection. Starting from an end-to-end adaptive sampling SNN detector (EAS-SNN), HAPQ conducts hardware-aware configuration search within discrete digital signal processor (DSP) and block RAM (BRAM) budgets, applies single-instruction-multiple-data (SIMD)-aligned structured pruning for computational regularity, and jointly quantizes synaptic weights and membrane potentials via a shift-friendly fixed-point recurrence. Evaluation on the Prophesee Gen1 dataset and an FPGA accelerator shows that HAPQ improves detection accuracy from 0.284 to 0.425 in mean average precision () and achieves 0.722 . Hardware implementation reveals a reduction in lookup table (LUT) usage to 1680, complete DSP elimination, and a maximum operating frequency of 920.81 MHz at 0.630 W. These results confirm that effective temporal SNN deployment requires joint optimization of model architecture, state precision, and hardware-aligned workload organization.

Purpose of research. Methods for embedding digital watermarks into digital objects constitute a scientific field at the intersection of information security and digital signal processing. Digital watermarks serve to ensure the authentication of digital objects by being inseparably embedded into them through the manipulation of data elements that constitute the digital object. A key requirement for digital watermarking methods is robustness, which entails that a watermark remains intact even after transmission over a heavily noisy channel. Therefore, the purpose of the research is to evaluate the robustness of watermarks to noise arising from the implementation of neural network removal attacks using the example of a watermark embedding algorithm that is robust to classical laboratory attacks. Methods. The study was conducted using the example of an algorithm for embedding a digital watermark into the coefficients of the discrete cosine transform of digital images using metaheuristic optimization. To implement a BlackBox removal attack, three neural network models were trained: one based on a convolutional neural network, a recurrent neural network, and a multilayer perceptron. Results. According to the results of experiments aimed at removing watermarks from the watermarked images as part of the BlackBox attack, the convolutional neural network architecture demonstrated the greatest efficiency. The watermarking algorithm demonstrated resistance to watermark removal by two other neural network models. Conclusion. The conducted research showed that the resilience of a digital watermark embedding algorithm to classical attacks does not guarantee resilience to neural network-based removal attacks. Resilience to this class of attacks must be incorporated at the stage of embedding algorithm synthesis. This can be achieved by reducing the determinism in the selection of watermark embedding areas in the image and by implementing anti-attack techniques at the embedding optimization stage.

Recent advancements in deep learning have shown considerable potential to enhance radar target detection, particularly in improving detection probability under complex environmental conditions. However, existing deep learning approaches largely operate in the real number domain, neglecting the complex-valued nature of radar data, and often inherit vision-oriented architectures that fail to address radar-specific challenges—such as sparse target echoes, the necessity for phase preservation, and constraints imposed by scanning radar systems. Meanwhile, conventional radar signal processing methods, including CA-CFAR, are limited by their dependence on idealized statistical models and often underperform in dynamic and cluttered electromagnetic environments.To overcome these issues, this paper proposes Radar Transformer for Detection (RaTDet), an end-to-end detection network that integrates complex-valued convolutional neural networks (CNNs) and Transformers. RaTDet fully leverages complex-valued data to preserve critical phase and amplitude information, enabling automated feature learning directly from raw radar signals. The model operates effectively with very few pulses, making it suitable for resource-constrained scenarios, and can serve as a pre-trained foundation model for various radar downstream tasks. Experimental results demonstrate that RaTDet achieves excellent detection performance, characterized by high detection probability (Pd) and low false alarm rate (Pfa), outperforming both traditional signal processing and conventional deep learning methods. This work bridges the gap between deep learning and radar signal processing, offering a flexible and powerful network for next-generation radar systems.