Deep Task-Based Quantization

Quantizers play a critical role in digital signal processing systems. Recent works have shown that the performance of acquiring multiple analog signals using scalar analog-to-digital converters (ADCs) can be significantly improved by prior to quantization. However, design such hybrid quantizers is quite complex, and their implementation requires complete knowledge statistical model signal. In this work we data-driven task-oriented quantization systems with ADCs, which determine mapping deep learning tools. These mappings are designed facilitate task recovering underlying information from quantized signals. By learning, circumvent need explicitly recover system find proper rule for it. Our main target application multiple-input multiple-output (MIMO) communication receivers, simultaneously acquire set signals, commonly subject constraints on number bits. results indicate that, MIMO channel estimation setup, proposed task-bask quantizer capable approaching optimal limits dictated indirect rate-distortion theory, achievable vector requiring model. Furthermore, symbol detection scenario, it demonstrated approach realize reliable bit-efficient receivers setting light task.