SEOFP-NET: Compression and Acceleration of Deep Neural Networks for Speech Enhancement Using Sign-Exponent-Only Floating-Points
Journal
IEEE/ACM Transactions on Audio Speech and Language Processing
Journal Volume
30
Pages
1016-1031
Date Issued
2021
Author(s)
Abstract
Numerous compression and acceleration strategies have achieved outstanding results on classification tasks in various fields, such as computer vision and speech signal processing. Nevertheless, the same strategies have yielded ungratified performance on regression tasks because the nature between these and classification tasks differs. In this paper, a novel sign-exponent- only floating-point network (SEOFP-NET) technique is proposed to compress the model size and accelerate the inference time for speech enhancement, a regression task of speech signal processing. The proposed method compressed the sizes of deep neural network (DNN)-based speech enhancement models by quantizing the fraction bits of single-precision floating-point parameters during training. Before inference implementation, all parameters in the trained SEOFP-NET model are slightly adjusted to accelerate the inference time by replacing the floating-point multiplier with an integer-adder. For generalization, the SEOFP-NET technique is introduced to different speech enhancement tasks in speech signal processing with different model architectures under various corpora. The experimental results indicate that the size of SEOFP-NET models can be significantly compressed by up to 81.249% without noticeably downgrading their speech enhancement performance, and the inference time can be accelerated to 1.212x compared with the baseline models. The results also verify that the proposed SEOFP-NET can cooperate with other efficiency strategies to achieve a synergy effect for model compression. In addition, the just noticeable difference (JND) was applied to the user study experiment to statistically analyze the effect of speech enhancement on listening. The results indicate that the listeners cannot facilely differentiate between the enhanced speech signals processed by the baseline model and the proposed SEOFP-NET. To the best knowledge of the authors, this study is one of the first research works to substantially compress the size of DNN-based algorithms and reduce the inference time of speech enhancement simultaneously while maintaining satisfactory enhancement performance. The promising results suggest that the application of DNN-based speech enhancement algorithms with the proposed SEOFP-NET technique is more suitable to light-weight embedded devices. Author
Subjects
Computational modeling
deep neural network model compression
floating-point integer arithmetic circuit
inference acceleration
Inference algorithms
Noise measurement
Noise reduction
Signal processing algorithms
speech dereverberation
Speech enhancement
speech enhancement
Task analysis
Audio signal processing
Digital arithmetic
Inference engines
Speech communication
Timing circuits
Arithmetic circuit
Computational modelling
Deep neural network model compression
Floating points
Floating-point integer arithmetic circuit
Inference acceleration
Inference algorithm
Integer arithmetic
Model compression
Neural network model
Noise measurements
Speech dereverberation
Deep neural networks
SDGs
Other Subjects
Audio signal processing; Digital arithmetic; Inference engines; Speech communication; Speech enhancement; Timing circuits; Arithmetic circuit; Computational modelling; Deep neural network model compression; Floating points; Floating-point integer arithmetic circuit; Inference acceleration; Inference algorithm; Integer arithmetic; Model compression; Neural network model; Noise measurements; Signal processing algorithms; Speech dereverberation; Task analysis; Deep neural networks
Type
journal article