A Hardware-Oriented and Memory-Efficient Method for CTC Decoding
This work addresses memory efficiency for hardware-based CTC decoders, which is incremental as it optimizes existing beam search and dictionary compression for specific applications.
The paper tackles the problem of high memory usage in CTC decoding for sequence tasks like speech and text recognition by proposing a hardware-oriented, memory-efficient method that compresses storage requirements by factors of 29.49 for ASR and 17.95 for STR with minimal precision loss.
The Connectionist Temporal Classification (CTC) has achieved great success in sequence to sequence analysis tasks such as automatic speech recognition (ASR) and scene text recognition (STR). These applications can use the CTC objective function to train the recurrent neural networks (RNNs), and decode the outputs of RNNs during inference. While hardware architectures for RNNs have been studied, hardware-based CTCdecoders are desired for high-speed CTC-based inference systems. This paper, for the first time, provides a low-complexity and memory-efficient approach to build a CTC-decoder based on the beam search decoding. Firstly, we improve the beam search decoding algorithm to save the storage space. Secondly, we compress a dictionary (reduced from 26.02MB to 1.12MB) and use it as the language model. Meanwhile searching this dictionary is trivial. Finally, a fixed-point CTC-decoder for an English ASR and an STR task using the proposed method is implemented with C++ language. It is shown that the proposed method has little precision loss compared with its floating-point counterpart. Our experiments demonstrate the compression ratio of the storage required by the proposed beam search decoding algorithm are 29.49 (ASR) and 17.95 (STR).