Many-Core Chip Time-Frequency Transform Kernel Mapping
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional time-frequency transform devices, such as ASICs, have limited flexibility and can only implement fixed algorithms, requiring data input to be a power of 2 or a product of a prime and power of 2, leading to poor flexibility in time-frequency transform operations.
Innovation Solution
A signal processing method utilizing a many-core chip that determines a transform kernel matrix based on a time domain signal and transform type, maps this matrix to processing cores, and inputs the time domain signal to perform various time-frequency transforms, including DFT, DCT, and FFT, enabling flexible and efficient processing across any interval.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If conventional efficient time-frequency transform devices are used, then computation efficiency is improved, but flexibility deteriorates
Solution Approach 1:
The patent implements a universal time-frequency transform device that can perform multiple transform types (DFT, DCT, FFT, CZT) using a single architecture. The processing cores are configured to handle different transform algorithms through programmable control, eliminating the need for separate dedicated devices for each transform type while maintaining high computation efficiency.
Solution Approach 2:
The device employs dynamic configuration of processing cores that can be adaptively allocated and programmed for different transform types based on real-time requirements. The system dynamically adjusts the number and arrangement of active processing cores according to the specific transform needs, enabling flexible adaptation without sacrificing computational performance.
2Device complexity
If fixed algorithm devices are used, then device complexity is reduced, but adaptability deteriorates
Solution Approach 1:
The transform device is segmented into multiple independent processing cores that can be individually configured and controlled. Each core can be programmed to handle specific transform operations, allowing the system to adapt to different transform types by activating and configuring appropriate core segments without redesigning the entire device.
Solution Approach 2:
The system changes operational parameters such as the number of active processing cores, their configuration, and control signals to adapt to different transform types. By dynamically adjusting these parameters, the device maintains relatively simple hardware architecture while achieving high adaptability across multiple transform algorithms.
3Speed
If data input is restricted to power of 2 or product of prime and power of 2, then processing speed is improved, but flexibility deteriorates
Solution Approach 1:
The patent introduces an intermediary data processing mechanism that handles arbitrary interval inputs by preprocessing or postprocessing the data to compatible formats for the processing cores. This intermediary layer enables the system to accept any interval length while maintaining efficient processing speeds by optimizing the data flow between the input and the core processing units.
Solution Approach 2:
The system dynamically adjusts the configuration of processing cores and data processing pathways based on the input interval length. Rather than being fixed to specific interval types, the device adapts its internal processing structure in real-time to handle any interval, maintaining high processing speeds through dynamic optimization of the computation path.
Data Source
AI summary
Provided are a signal processing method based on a many-core chip, an electronic device and a medium. The method includes: determining, according to a time domain signal to be processed and a time-frequency transform type of the time domain signal, a transform kernel matrix of the time domain signal; mapping the transform kernel matrix to a plurality of processing cores of the many-core chip; and mapping the time domain signal to the plurality of processing cores so that the plurality of processing cores determine, according to the transform kernel matrix and the time domain signal, a frequency domain signal corresponding to the time domain signal.