21 October 2008 in Journal of Real-Time Image Processing
25 July 2017 in IEEE Access
MIMohammed Shaaban Ibraheem
Mohammed Shaaban Ibraheem
An embedded real-time indexing engine for live radio stations requires a high-speed parallel advanced audio coding (AAC) decoder architecture to decode hundreds of compressed audio streams broadcast in the digital radio mondiale band. Several AAC hardware core units must be integrated into a single chip and synchronized with a global controller to achieve such high performance. This paper proposes a new parallel AAC decoder architecture to address this challenge. The proposed architecture includes multiple AAC decoder core units, each of which achieves a speed-up by a factor of 2 compared with the existing AAC decoder core units while using optimal logic resources. The proposed architecture overcomes several challenges faced by existing architectures. The Huffman decoder module decodes one word per clock cycle regardless of word length, and a smaller lookup table size is achieved for the inverse quantization module. The inverse modified discrete cosine transform architecture is fully pipelined, and the resource sharing technique is used to reduce the logic area. An initial prototype is implemented on an FPGA platform.
26 September 2024 in ACM Transactions on Embedded Computing Systems
Embedded systems based on Microcontroller Units (MCUs) often gather significant quantities of data and solve various issues. Convolutional Neural Networks (CNNs) have proven their effectiveness in solving computer vision and natural language processing tasks. However, implementing CNNs within MCUs is challenging due to their high inference costs, which varies widely depending on hardware targets and CNN topologies. Despite state-of-the-art advancements, no efficient design space exploration solutions handle the wide variety of implementation solutions. In this article, we introduce the ZIP-CNN design space exploration methodology, which facilitates CNN implementation within MCUs. We developed a model that quantitatively estimates the latency, energy consumption, and memory space required to run a CNN within an MCU. This model accounts for algorithmic reductions such as knowledge distillation, pruning, or quantization and applies to any CNN topology. To demonstrate the efficiency of our methodology, we investigated LeNet5, ResNet8, and ResNet26 within three different MCUs. We made materials and supplementary results available in a GitHub repository: https://github.com/ThGbay/ZIP-CNN . The proposed method was empirically verified on three hardware targets running at 14 different operating frequencies. The three CNN topologies investigated were implemented in their default configuration in FP32, and also reduced with INT8 quantization, pruning at five different rates and with knowledge distillation. The estimates of our model are very reliable with an error of 3.29% to 15.23% for latency, 3.12% to 10.34% for energy consumption, and 1.95% to 6.31% for memory space. These results are based on on-device measurements.