基于深度神經(jīng)網(wǎng)絡(luò)的Morse碼自動譯碼算法

游凌; 李偉浩; 張文林; 王科人

doi:10.11999/JEIT190658

基于深度神經(jīng)網(wǎng)絡(luò)的Morse碼自動譯碼算法

doi: 10.11999/JEIT190658 cstr: 32379.14.JEIT190658

游凌²,
李偉浩^{1, 2, ,},
張文林¹,
王科人²

1.
戰(zhàn)略支援部隊信息工程大學(xué) 鄭州 450001
2.
盲信號處理國家級重點實驗室成都 610041

基金項目: 國家自然科學(xué)基金(61403415)，中國博士后科學(xué)基金(2016M602975)

詳細信息

作者簡介:
游凌：男，1971年生，博士，研究員，研究方向為信號分析

李偉浩：男，1996年生，碩士生，研究方向為深度學(xué)習(xí)、信號分析

張文林：男，1982年生，博士，副教授，研究方向為語音信號處理、語音識別、自然語言理解

王科人：男，1986年生，博士，助理研究員，研究方向為信號分析、智能信息處理

通訊作者:
李偉浩　liweihao315@gmail.com

中圖分類號: TN919.32
計量
- 文章訪問數(shù): 4180
- HTML全文瀏覽量: 1263
- PDF下載量: 190
- 被引次數(shù): 0
出版歷程
- 收稿日期: 2019-08-29
- 修回日期: 2020-05-08
- 網(wǎng)絡(luò)出版日期: 2020-05-28
- 刊出日期: 2020-11-16

Automatic Decoding Algorithm of Morse Code Based on Deep Neural Network

Ling YOU²,
Weihao LI^{1, 2
, ,},
Wenlin ZHANG¹,
Keren WANG²

1.
PLA Strategic Support Force Information Engineering University, Zhengzhou 450001, China
2.
National Key Laboratory of Science and Technology on Blind Signal Processing, Chengdu 610041, China

Funds: The National Natural Science Foundation of China (61403415), The Postdoctoral Science Foundation of China(2016M602975)

摘要

摘要: 在軍用和民用領(lǐng)域，Morse電報一直是一種重要的短波通信手段，但目前的自動譯碼算法仍然存在準確率低、無法適應(yīng)低信噪比和不穩(wěn)定的信號等問題。該文引入深度學(xué)習(xí)方法構(gòu)建了一個Morse碼自動識別系統(tǒng)，神經(jīng)網(wǎng)絡(luò)模型由卷積神經(jīng)網(wǎng)絡(luò)、雙向長短時記憶網(wǎng)絡(luò)和連接時序分類層組成，結(jié)構(gòu)簡單，且能夠?qū)崿F(xiàn)端到端的訓(xùn)練。相關(guān)實驗表明，該譯碼系統(tǒng)在不同信噪比、不同碼速、信號出現(xiàn)頻率漂移以及不同發(fā)報手法引起的碼長偏差等情況下，均能取得較好的識別效果，性能優(yōu)于傳統(tǒng)的自動識別算法。
- Morse碼 /
- 自動譯碼 /
- 深度學(xué)習(xí) /
- 頻率漂移 /
- 碼長偏差
Abstract: In the military and civilian fields, the Morse telegraph is always as an important means of short-wave communication, but the current automatic decoding algorithms still have problems such as low accuracy, inability to adapt to low signal-to-noise ratio and unstable signals. A deep learning method is introduced to construct a Morse code automatic recognition system. The neural network model consists of convolutional neural network, bidirectional long short-term memory network and connectionist temporal classification layer. The structure is simple and can implement end-to-end training. Related experiments show that the decoding system can achieve good recognition results under different signal-to-noise ratio, code rate, frequency drift and code length deviation caused by different sending manipulation, and the performance is better than the traditional recognition algorithms.
- Morse code /
- Automatic decoding /
- Deep learning /
- Frequency drift /
- Code length deviation

HTML全文

圖 1 Morse信號時頻圖

下載: 全尺寸圖片幻燈片

圖 2 神經(jīng)網(wǎng)絡(luò)結(jié)構(gòu)

下載: 全尺寸圖片幻燈片

圖 3 特征序列與時頻圖的對應(yīng)

下載: 全尺寸圖片幻燈片

圖 4 多模型識別準確率對比

下載: 全尺寸圖片幻燈片

圖 5 多模型識別速度對比

下載: 全尺寸圖片幻燈片

圖 6 出現(xiàn)頻率漂移和碼長偏差的Morse信號

下載: 全尺寸圖片幻燈片

表 1 CNN層設(shè)置

層名稱	對應(yīng)核大小
卷積層1	(5, 5, 1, 32)，步長=(1, 1)
最大池化層1	(2, 2)，步長=(2, 2)
卷積層2	(5, 5, 32, 64)，步長=(1, 1)
最大池化層2	(2, 16)，步長=(2, 2)

下載: 導(dǎo)出CSV

表 2 數(shù)據(jù)集組成

	碼速(wpm)	信噪比(dB)	數(shù)目
訓(xùn)練集	25, 30, 40	40, 30, 20, 10, 6, 3, –3, –6, –8, –10	25000/50000
驗證集	25, 30, 40	40, 30, 20, 10, 6, 3, –3, –6, –8, –10	2500
測試集	25, 30, 40	40, 30, 20, 10, 6, 3, –3, –6, –8, –10	2500

下載: 導(dǎo)出CSV

表 3 頻率漂移和碼長偏差情況下的譯碼準確率

	字準確率(%)	詞準確率(%)
原始信號	99.92	99.65
頻率漂移	96.23	91.71
頻率漂移+碼長偏差	95.88	90.40

下載: 導(dǎo)出CSV

表 4 有無頻漂時去掉CNN前后譯碼性能

	迭代次數(shù)	字準確率(%)	詞準確率(%)
有CNN，無頻漂	23	99.92	99.65
無CNN，無頻漂	42	92.71	73.90
有CNN，有頻漂	26	96.23	91.71
無CNN，有頻漂	47	63.11	20.35

下載: 導(dǎo)出CSV

參考文獻(17)

DEY S, CHUGG K M, and BEEREL P A. Morse code datasets for machine learning[C]. The 9th International Conference on Computing, Communication and Networking Technologies (ICCCNT), Bangalore, India, 2018: 1–7. doi: 10.1109/icccnt.2018.8494011.

SHIH C H and LUO C H. A Morse-code recognition system with LMS and matching algorithms for persons with disabilities[J]. International Journal of Medical Informatics, 1997, 44(3): 193–202. doi: 10.1016/s1386-5056(97)00020-8

HSIEH M C, LUO C H, and MAO Chiwu. Unstable Morse code recognition with adaptive variable-ratio threshold prediction for physically disabled persons[J]. IEEE Transactions on Rehabilitation Engineering, 2000, 8(3): 405–413. doi: 10.1109/86.867882

YANG Chenghong, LUO C H, JEANG Y L, et al. A novel approach to adaptive Morse code recognition for disabled persons[J]. Mathematics and Computers in Simulation, 2000, 54(1/3): 23–32. doi: 10.1016/s0378-4754(00)00180-4

GOLD B. Machine recognition of hand-sent Morse code[J]. IRE Transactions on Information Theory, 1959, 5(1): 17–24. doi: 10.1109/TIT.1959.1057478

WU Chungmin and LUO Chinghsing. Morse code recognition system with fuzzy algorithm for disabled persons[J]. Journal of Medical Engineering & Technology, 2002, 26(5): 202–207. doi: 10.1080/03091900210156904

WANG Yaqi, SUN Zhonghua, and JIA Kebin. An automatic decoding method for Morse signal based on clustering algorithm[C]. The 12th International Conference on Intelligent Information Hiding and Multimedia Signal Processing, Kaohsiung, China, 2017: 235–242.

王亞琦, 孫中華, 賈克斌. Morse報自動譯碼算法研究[J]. 信號處理, 2017, 33(11): 1451–1456.

WANG Yaqi, SUN Zhonghua, and JIA Kebin. Research on automatic decoding algorithm of Morse telegraph[J]. Signal Processing, 2017, 33(11): 1451–1456.

LECUN Y, BENGIO Y, and HINTON G. Deep learning[J]. Nature, 2015, 521(7553): 436–444. doi: 10.1038/nature14539

WANG Xianyu, ZHAO Qi, MA Cheng, et al. Automatic Morse code recognition under low SNR[C]. 2018 International Conference on Mechanical, Electronic, Control and Automation Engineering (MECAE 2018), Qingdao, China, 2018: 46–51. doi: 10.2991/mecae-18.2018.46.

SHI Baoguang, BAI Xiang, and YAO Cong. An end-to-end trainable neural network for image-based sequence recognition and its application to scene text recognition[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2017, 39(11): 2298–2304. doi: 10.1109/TPAMI.2016.2646371

韋醒超. 摩爾斯短波無線通信系統(tǒng)的設(shè)計與實現(xiàn)研究[D]. [碩士論文], 湖南大學(xué), 2017.

WEI Xingchao. Research on the design and implementation of Morse shortwave wireless communication system[D]. [Master dissertation], Hunan University, 2017.

CABAL-YEPEZ E, GARCIA-RAMIREZ A G, ROMERO-TRONCOSO R J, et al. Reconfigurable monitoring system for time-frequency analysis on industrial equipment through STFT and DWT[J]. IEEE Transactions on Industrial Informatics, 2013, 9(2): 760–771. doi: 10.1109/TII.2012.2221131

SAINATH T N, KINGSBURY B, SAON G, et al. Deep convolutional neural networks for large-scale speech tasks[J]. Neural Networks, 2015, 64: 39–48. doi: 10.1016/j.neunet.2014.08.005

GRAVES A and SCHMIDHUBER J. Framewise phoneme classification with bidirectional LSTM and other neural network architectures[J]. Neural Networks, 2005, 18(5/6): 602–610. doi: 10.1016/j.neunet.2005.06.042

GRAVES A, FERNáNDEZ S, and GOMEZ F. Connectionist temporal classification: Labelling unsegmented sequence data with recurrent neural networks[C]. The 23rd International Conference on Machine Learning, Pittsburgh, USA, 2006: 369–367.

劉宏哲, 楊少鵬, 袁家政, 等. 基于單一神經(jīng)網(wǎng)絡(luò)的多尺度人臉檢測[J]. 電子與信息學(xué)報, 2018, 40(11): 2598–2605. doi: 10.11999/JEIT180163

LIU Hongzhe, YANG Shaopeng, YUAN Jiazheng, et al. Multi-scale face detection based on single neural network[J]. Journal of Electronics &Information Technology, 2018, 40(11): 2598–2605. doi: 10.11999/JEIT180163

相關(guān)文章

施引文獻

資源附件(0)

訪問統(tǒng)計