⚡ electrical engineering

GNU Radio-Based Performance Evaluation of Confidence-Based Soft-decision Recovery for Noisy BPSK Communication

This paper presents a GNU Radio-based evaluation demonstrating that a confidence-based soft-decision recovery method utilizing repeated bit transmission and real-valued samples significantly outperforms conventional hard-decision majority voting in reducing bit errors for noisy BPSK communication.

Original authors: Jenifer Darling Rosita Prince Devaraj, Stalin Jacob Wilson Packia Dhas

Published 2026-07-15

📖 5 min read🧠 Deep dive

CC BY 4.0

Original authors: Jenifer Darling Rosita Prince Devaraj, Stalin Jacob Wilson Packia Dhas

Original paper licensed under CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

Imagine you are trying to send a secret message to a friend across a very noisy room. You can only shout "Zero" or "One," but the wind is howling, and sometimes your friend hears the wrong thing. This is exactly the problem engineers face with digital communication, and this paper is like a detective story about how to fix a broken message in that noisy room.

The Old Way: The "Shout and Guess" Game
Usually, when a computer receives a signal, it uses a simple rule called a "Binary Slicer." Think of this like a strict referee who listens to your shout and immediately decides: "If it sounds even a little bit like 'One,' it's a One. If it sounds like 'Zero,' it's a Zero." The referee doesn't care how loud or clear the shout was; they just pick a side.

To make things safer, the sender repeats every single bit of the message seven times. So, if you want to send a "One," you shout "One-One-One-One-One-One-One." The referee then uses "majority voting." If four or more of the shouts sound like "One," the referee writes down "One."

The paper argues that this old way is a bit too rigid. It throws away useful clues. If the referee hears a shout that is barely louder than the wind, they still treat it exactly the same as a shout that was booming and clear. They lose the "confidence" information.

The New Way: The "Confidence Score" Detective
The authors, Jenifer Darling Rosita Prince Devaraj and Stalin Jacob Wilson Packia Dhas, tested a smarter approach using a tool called GNU Radio (which is like a giant, flexible Lego set for building radio experiments). Instead of just listening to the final "One" or "Zero" decision, their new method listens to the actual sound wave before the referee makes a decision.

Imagine the referee now has a "confidence meter."

The Summation Trick: Instead of counting votes, they add up the actual strength of all seven shouts. If the total "push" of the sound waves is positive, they decide it's a "One." If it's negative, it's a "Zero."
The Confidence Score: They also calculate a score between 0 and 1 to see how sure they are. If the seven shouts were all very loud and clear, the score is high. If they were weak and wavering near the edge of the noise, the score is low.

What They Found (The Results)
The researchers tested this with three different 24-bit messages (short secret codes) and cranked up the "noise voltage" to simulate a very stormy room. They went up to 5.0 V of noise.

Here is the big reveal: The new "confidence" method was better at fixing the message than the old "majority voting" method in almost every case.

The Numbers: In one specific test case (Input 3) at 5.0 V of noise, the old method made 12 mistakes out of the 24 bits (a Bit Error Rate, or BER, of 0.5000). The new method only made 5 mistakes (a BER of 0.2083).
The Zero-Error Zone: For some messages, the new method got zero errors even when the noise was high enough to make the old method fail completely. For example, with Input 2, the new method had zero errors up to 2.5 V of noise, while the old method was already making mistakes.
The "Weak Bit" Bonus: The best part of the new method is that it can point a finger at the bits it isn't sure about. If a bit has a confidence score below 0.35, the system flags it as "weak" or "uncertain." It's like the referee saying, "I think this is a One, but I'm not totally sure, so check this one again."

What This Means (and What It Doesn't)
The paper shows that using the real, wiggly sound waves (soft-decision) instead of just the final yes/no decision helps recover messages better in noisy conditions. It proves that keeping track of "how loud" the signal was makes a difference.

However, the authors are careful to say this isn't a magic wand that fixes everything. They note that while it improved things in most cases, it didn't guarantee zero errors for every single test condition. They also mention that this was a simulation and experiment using specific software (GNU Radio) and Python, not a real-world satellite launch just yet.

The Bottom Line
If you are trying to send a short, important message through a noisy channel, this paper suggests that listening to the strength of the signal and keeping a "confidence score" is a smarter move than just counting votes. It reduces errors and tells you exactly which parts of your message might need a second look. It's a low-complexity way to make communication more robust, perfect for future experiments or even for building systems that need to hide messages in plain sight.

Technical Summary: GNU Radio-Based Performance Evaluation of Confidence-Based Soft-Decision Recovery for Noisy BPSK Communication

Problem Statement
In noisy digital communication systems, accurate bit recovery is critical, particularly when short binary messages must be transmitted through channels subject to additive noise and distortion. While Binary Phase Shift Keying (BPSK) is a robust modulation scheme widely used in satellite navigation and baseline performance studies, high channel noise can cause received symbols to cross detection thresholds, leading to bit errors. Conventional hard-decision receivers, such as the Binary Slicer, convert received signal samples directly into binary 0s and 1s. Although simple, this approach discards valuable reliability information contained in the signal amplitude; samples near the decision threshold and those far from it are treated identically once sliced. While repetition coding combined with hard-decision majority voting offers a basic error recovery mechanism, it relies solely on final binary decisions, failing to utilize the confidence levels inherent in real-valued received samples.

Methodology
The authors implemented a performance evaluation framework using GNU Radio and Python to compare a proposed confidence-based soft-decision recovery method against a baseline hard-decision Binary Slicer majority voting approach.

System Model: The experimental setup utilized a GNU Radio flowgraph to model a BPSK transmitter, a noisy channel (using a Channel Model block with varying noise voltage levels), and a receiver. The transmitter mapped binary bits to BPSK symbols (0 to -1, 1 to +1).
Frame Structure: Transmitted data consisted of structured frames containing a 6-bit preamble, a 24-bit message, repetition-coded message bits, and 3 padding bits. Each of the 24 message bits was repeated seven times, resulting in a 168-bit message section and a total frame length of 177 bits.
Baseline Method (Hard-Decision): The received signal was passed through a Binary Slicer to generate hard binary outputs. The receiver then applied majority voting to groups of seven bits to recover the original message.
Proposed Method (Soft-Decision): Instead of slicing immediately, the system captured real-valued BPSK samples. For each group of seven samples corresponding to a single message bit, the method summed the real-valued samples. The sign of the group sum determined the recovered bit (positive/zero $\rightarrow$ 1, negative $\rightarrow$ 0).
Confidence Estimation: A confidence value ( $C_i$ ) was calculated for each recovered bit by dividing the absolute value of the group sum by the sum of the absolute values of the individual samples in that group. Bits with confidence values below a threshold of 0.35 were flagged as "weak" or uncertain.
Evaluation Metrics: Performance was assessed using the number of bit errors, Bit Error Rate (BER), and the identification of weak-bit positions. Tests were conducted across three different 24-bit input messages and noise voltage levels ranging from 0.5 V to 5.0 V.

Key Contributions
The primary contribution of this work is the experimental evaluation of a confidence-based soft-decision repetition recovery method specifically for noisy BPSK bit detection. The study provides a direct comparison between this soft-decision approach and traditional hard-decision majority recovery. It demonstrates how utilizing real-valued signal samples before hard slicing can reduce bit errors and provide reliability information through weak-bit identification. The implementation serves as a low-complexity recovery layer suitable for robust, low-data-rate communication experiments.

Results
Experimental results indicated that the confidence-based soft-decision method consistently outperformed the raw Binary Slicer majority approach in most tested scenarios:

Error Reduction: Across various noise levels and input messages, the soft-decision method reduced the number of bit errors. For instance, at 5.0 V noise voltage with Input 3, the error count dropped from 12 (hard-decision) to 5 (soft-decision), reducing the BER from 0.5000 to 0.2083.
Zero-Error Cases: For certain inputs, the proposed method achieved zero bit errors at noise voltage levels where the hard-decision method produced multiple errors (e.g., Input 2 at 1.0 V and 2.5 V).
Weak-Bit Identification: The method successfully identified uncertain bit positions using the 0.35 confidence threshold, offering additional metadata about the reliability of the recovered message that the hard-decision method lacks.
BER Trends: As noise voltage increased, the BER for the hard-decision method rose sharply, while the soft-decision method maintained a significantly lower BER, with improvements becoming more pronounced at higher noise levels.

Significance and Claims
The paper claims that confidence-based soft-decision recovery effectively improves noisy BPSK bit detection by leveraging real-valued signal information that is typically discarded by hard slicing. The method provides a practical, low-complexity solution for improving message recovery in low-data-rate experiments. The authors note that while the method improved performance in most cases, it did not guarantee error reduction for every individual test condition. The work is positioned as a foundation for future extensions, including the optimization of confidence thresholds, testing under additional channel conditions, and integration with stronger error-control coding or end-to-end deep learning frameworks for robust communication under realistic channel conditions. The authors also suggest potential utility for improving the reliability of underlying bit recovery stages in robust covert or steganographic communication systems.

Technical Summary: GNU Radio-Based Performance Evaluation of Confidence-Based Soft-Decision Recovery for Noisy BPSK Communication

More like this