Abstract
Resolving signal collisions in wireless networks typically requires multiple antennas or pre-assigned codes. This paper presents a novel blind source separation (BSS) approach that enables a single-antenna receiver to untangle two overlapping signals by combining oversampling with an optimized pulse-shape design. By maximizing the pulse's time-domain decay rate through convex optimization, the system creates a well-conditioned "virtual MIMO" environment. Simulations demonstrate that the proposed method achieves a 4 dB SNR advantage over standard IOTA pulses while maintaining spectral mask compliance.
Introduction
In random-access wireless networks, collisions—where multiple users transmit simultaneously on the same frequency—are a primary bottleneck for throughput. Conventional solutions like spatial diversity (multiple antennas) or code diversity (CDMA) often impose hardware costs or bandwidth overhead. This research addresses the challenging scenario of two users transmitting with equal power to a single-antenna receiver without pre-assigned codes. The goal is to transform this collision into a solvable mathematical problem using only signal processing at the receiver.
Methodology: Virtual MIMO and Pulse Optimization
The proposed strategy relies on three integrated components:
1. Virtual MIMO via Oversampling
By sampling the received signal at a rate higher than the symbol rate, the receiver generates multiple polyphase components. These components serve as "virtual antennas," providing the mathematical diversity necessary to separate the mixed signals.
2. Optimal Pulse-Shape Design
The core innovation lies in the pulse-shape engineering. Standard pulses like the Raised Cosine or IOTA are designed for spectral efficiency or zero inter-symbol interference (ISI). However, for BSS, the pulse must also minimize correlation between virtual channels. The authors formulate a convex optimization problem using semidefinite relaxation to design a pulse that decays as rapidly as possible in the time domain.
As shown in the comparison above, the optimized pulse is "sharper" than the IOTA benchmark. This characteristic is achieved by maximizing the energy of the pulse's first derivative, which directly improves the condition number of the channel matrix, making the signals easier to pull apart.
3. Intentional Delay and SIC
The system introduces a half-symbol delay ($T_s/2$) between users. After initial separation using the Joint Approximate Diagonalization of Eigenmatrices (JADE) algorithm, Successive Interference Cancellation (SIC) is applied to refine the recovery of the weaker signal.
Key Results
The performance of the proposed pulse was evaluated using 4-QAM modulated signals in a flat-fading environment.
Signal Recovery Performance
The most significant finding is the improvement in Symbol Error Rate (SER). At a symbol rate of 10 M/sec, the proposed pulse provides an approximate 4 dB SNR advantage over the IOTA pulse to achieve the same SER level.
Spectral Compliance
Despite its rapid time-domain decay, the optimized pulse remains strictly compliant with the IEEE 802.11 spectral mask. In contrast, standard pulses like IOTA were found to violate the mask at higher symbol rates (e.g., 12.19 M/sec), highlighting the robustness of the proposed design.
Conclusion
This research demonstrates that clever pulse-shape engineering can substitute for physical hardware diversity in resolving wireless collisions. By treating the pulse shape as a tool for channel conditioning rather than just a spectral filter, the authors enable robust blind signal separation in single-antenna systems. Future work will explore scaling this approach to more than two users and integrating it into next-generation MAC protocols.
Summary updated