End to End Deep Neural Network Frequency Demodulation of Speech Signals

Elbaz, Dan; Zibulevsky, Michael

doi:10.1007/978-3-030-03402-3_1

Cited by 10 publications

(9 citation statements)

References 20 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In the following, the demodulation performance evaluation experiments are carried out for the three signals respectively. At the same time, in order to highlight the performance advantages of DeepDeFM proposed by us, the Bi-LSTM network mentioned in [28] is added for comparison.…”

Section: B Effect Of Noisementioning

confidence: 99%

“…Zhang et al constructed a DNN-based demodulator with LSTM unit assistance for orthogonal frequency division multiplexingaided differential chaos shift keying (OFDM-DCSK) systems in [27]. In [28], the authors adopted the method based on end-to-end learning, and added the bidirectional-LSTM (Bi-LSTM) network to demodulate FM signal, which can better reconstruct the speech signal.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Deep Learning-Based FM Demodulation in Complex Electromagnetic Environment

Zheng,

Pei,

Chen

et al. 2024

IEEE Open J. Commun. Soc.

View full text Add to dashboard Cite

In recent years, deep learning has been applied widely in the field of communication. In this paper, a deep learning-based frequency modulation (FM) demodulation method is proposed for FM signal demodulation in complex electromagnetic environment. The received in-phase and quadrature (IQ) signals are fed into a designed model to recover the source signal. We perform experiments with random, voice and broadcasting signals under noise, multipath fading and single tone jamming scenarios. The results demonstrate that the proposed method outperforms the traditional method in terms of demodulation performance across all scenarios. Especially in the case of multipath fading and single tone jamming, our proposed method exhibits significant performance improvements over traditional demodulation. We further propose to use downsampling to reduce the computational complexity of DeepDeFM. The results indicate that by selecting a proper downsampling factor, it is possible to significantly reduce time complexity while maintaining high sound quality with minimal attenuation.

show abstract

Section: B Effect Of Noisementioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Deep Learning-Based FM Demodulation in Complex Electromagnetic Environment

Zheng,

Pei,

Chen

et al. 2024

IEEE Open J. Commun. Soc.

View full text Add to dashboard Cite

show abstract

“…(Ramjee et al, 2019), (J. H. Lee, Kim, Kim, Yoon, & Choi, 2017), (Zhang et al, 2018), (Chikha, Dayoub, Hamouda, & Attia, 2014), (Yashashwi, Sethi, & Chaporkar, 2019), (Mendis, Wei, & Madanayake, 2019), (Liu, Yang, & Gamal, 2017) Sinyal Tanıma (H. , (U. Mohammad & Sorour, 2018), (Elbaz & Zibulevsky, 2018), (Chen & Laneman, 2006) (Karanov, Lavery, Bayvel, & Schmalen, 2019), (He, Wen, Jin, & Li, 2018a), (Y. Yang, Gao, Ma, & Zhang, 2019), (Hao Ye, Li, & Juang, 2018)Ye, (Soltani, Pourahmadi, Mirzaei, & Sheikhzadeh, 2019), (Arnold et al, 2019), (Jiang et al, 2018), (Jiang et al, 2019), (Cheng, Liu, Wang, Yan, & Zhu, 2019), (Fujihashi, Koike-Akino, Watanabe, & Orlik, 2018), (H. Ye & Li, 2017), (Kang, Chun, & Kim, 2018),(T. J. O 'Shea, Erpek, & Clancy, n.d.), (Hao Ye et al, 2018), (Y.…”

Section: Derin öğRenmementioning

confidence: 99%

Haberleşme Sistemlerinde Derin Öğrenme

Yildirim¹

2020

European Journal of Science and Technology

View full text Add to dashboard Cite

Deep learning has become the most successful learning method in machine learning. While deep learning provides a clear advantage over other machine learning methods, especially when the amount of data is high, it can produce an approximate result to other machine teaching methods when data is low. This new learning method has the potential to contribute to many innovations, from redesigning the physical layers used in communication technologies to modeling wireless networks. It is particularly useful in communication systems where mathematical modeling is difficult, for example, 5G and molecular communication. Therefore, many types of research on the application of deep learning in communication systems have been conducted recently. However, the distance of institutions and researchers about communication technologies to deep learning methods has limited the number and impact of these studies. Therefore, it is necessary to collectively examine the studies that involve the application of deep learning to communication technologies, to evaluate the achievements, and to contribute to the determination of new research topics. For this purpose, in this study, firstly, the achievements of deep learning and usage areas are summarized and then the studies that contribute to the development of communication technologies are classified and examined comparatively. To make deep learning more effective in communication, what needs to be done were discussed and deep learning based research areas that could lead to next generation communication systems were determined.

show abstract

“…In addition, Elbaz et al. (2018) proposed that using the long and short memory network (LSTM) to learn mapping directly from baseband modulated signals to modulated audio at the transmitter, which would create baseband to speech mappings. This method is proven to perform better than the traditional receiving system at low signal‐to‐noise ratio (SNR) conditions.…”

Section: Introductionmentioning

confidence: 99%

GMSK Demodulation Combining 1D‐CNN and Bi‐LSTM Network Over Strong Solar Wind Turbulence Channel

Zhu

Yang

et al. 2023

Radio Science

View full text Add to dashboard Cite

During the deep space exploration of the solar system, the reliable communication link is vital, since it is the only way that enables the transmission of telecommand from the earth and the return of telemetry information and scientific data from the probe. However, the communication link in the solar system space environment faces many severe challenges. In addition, due to the ultralong communication distance, the communication link between the Earth and the solar system probe may encounter solar scintillation due to solar wind during near superior solar conjunction (Morabito & Hastrup, 2001). Solar scintillation is a telecommand and telemetry signal fluctuation phenomenon. When the transmitted telecommand and telemetry signals pass through the solar corona region, the solar wind causes telecommand and telemetry signal amplitude fluctuation, phase fluctuation and spectrum expansion. Especially, if the solar scintillation is sufficiently strong that means the heliocentric distance is less than four solar radii(Xu & Song, 2019), the communication link may be intermittently or even completely interrupted, which results in communication failure

show abstract

End to End Deep Neural Network Frequency Demodulation of Speech Signals

Cited by 10 publications

References 20 publications

Deep Learning-Based FM Demodulation in Complex Electromagnetic Environment

Deep Learning-Based FM Demodulation in Complex Electromagnetic Environment

Haberleşme Sistemlerinde Derin Öğrenme

GMSK Demodulation Combining 1D‐CNN and Bi‐LSTM Network Over Strong Solar Wind Turbulence Channel

Contact Info

Product

Resources

About