Multi-Symbol Detection via Hypothesis Subtraction and Metric Optimization
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In mobile wireless communication systems, particularly in 3G systems using CDMA codes, the detection of digital data symbols is hindered by multipath propagation and non-orthogonal symbol waveforms, leading to increased computational burden in signal processing due to the need to test numerous hypotheses for symbol detection.
Innovation Solution
A method that hypothesizes all but one symbol, subtracts their effects from mixed detection statistics, combines and quantizes the remainders, and computes a metric for each hypothesis to determine the best symbol set, reducing the number of necessary computations by using modified dependence coefficients and pre-computed values to efficiently decode symbols.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If brute-force detection is used to test all possible symbol combinations, then symbol detection accuracy is improved, but computational burden increases significantly
Solution Approach 1:
The patent segments the detection process into two stages: first hypothesizing all but one symbol to reduce the search space, then determining the remaining symbol through subtraction and quantization. This segmentation reduces computational complexity from testing all M^N combinations to testing only M^(N-1) combinations, while maintaining detection accuracy through the metric computation step that evaluates the complete symbol set.
Solution Approach 2:
The patent performs preliminary action by hypothesizing all but one symbol before final detection. This preliminary hypothesis reduces the number of computations required in the main detection process. The hypothesized symbols are used to subtract their effects from the mixed statistics, leaving a simplified expression that can be directly quantized to determine the remaining symbol with minimal additional computation.
2Productivity
If multiple data streams are assigned to increase data rate, then throughput is improved, but signal separation becomes more difficult due to multipath propagation
Solution Approach 1:
The patent uses feedback through metric computation to evaluate the quality of detected symbols. The metric is computed based on the hypothesized symbols and the determined remaining symbol, providing feedback on how well the detected symbol set explains the received signal. This feedback mechanism enables accurate signal separation even in multipath environments by selecting the symbol set that minimizes the metric, effectively separating overlapping data streams despite channel distortions.
3Reliability
If orthogonal codes are used for signal separation, then interference cancellation is improved, but orthogonality is lost when codes are relatively delayed due to multipath
Solution Approach 1:
The patent changes the detection approach from relying on code orthogonality to using metric-based optimization. Instead of assuming orthogonal codes maintain separation, the patent computes detection statistics that capture the actual mixing of symbols due to multipath, then uses metric computation to find the symbol set that best explains the observed statistics. This parameter change from orthogonality-based separation to metric-based detection maintains reliability even when orthogonality is lost.
Data Source
AI summary
Where two or more multi-valued digital data symbols are modulated so that they overlap after passing through a channel, forming a combined signal, a receiver receives the combined signal and forms detection statistics to attempt to recover the symbols. Where forming detection statistics does not completely separate the symbols, each statistic comprises a different mix of the symbols. A receiver determines the symbols which, when mixed in the same way, reproduce or explain the statistics most closely. For example, the receiver hypothesizes all but one of the symbols and subtracts the effect of the hypothesized symbols from the mixed statistics. The remainders are combined and quantized to the nearest value of the remaining symbol. For each hypothesis, the remaining symbol is determined. A metric is then computed for each symbol hypothesis including the so-determined remaining symbol, and the symbol set producing the best metric is chosen as the decoded symbols.