A Convolutional Neural Network Smartphone App for Real-Time Voice Activity Detection

Sehgal, Abhishek; Kehtarnavaz, Nasser

doi:10.1109/access.2018.2800728

Cited by 99 publications

(50 citation statements)

References 20 publications

(27 reference statements)

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…With the rapid development of advanced communication technology, mobile wireless and Voice over IP (VoIP) are widely used around the world, and speech steganography has increasingly high value covering the secure and covert communication. Speech is a special case of audio signals, and it is different from the typical audio signals in terms of spectral bandwidth, intensity distribution, and signal continuity [4]- [6]. In general, methods designed for audio steganography are not suitable for speech steganography because those methods take the media object as continuous signal and do not consider the speech characteristics.…”

Section: Introductionmentioning

confidence: 99%

Robust Speech Steganography Using Differential SVD

et al. 2019

View full text Add to dashboard Cite

The speech signal is different from the typical audio in terms of spectral bandwidth, intensity distribution, and signal continuity, thus how to achieve high imperceptibility and strong robustness for speech steganography is a big challenge. In this paper, we present a speech steganography scheme based on the parity-segmented method and the differential singular value decomposition (SVD). The selected discrete cosine transform (DCT) coefficients are divided into two segments according to parity order. In this way, the energy of the paired segments is approximately equal, therefore the changes in the singular values caused by data embedding are reduced, and high imperceptibility is achieved. Unlike the common SVD-based steganography, the differential SVD scheme can effectively remove the impact of amplitude scaling attack by embedding the secret message into the difference between the singular values. Experimental results show that the proposed method achieves high imperceptibility and strong robustness while resisting the state-of-the-art steganalytic methods. INDEX TERMS Steganography, differential SVD, paired segments, imperceptibility, amplitude scaling.

show abstract

Section: Introductionmentioning

confidence: 99%

Robust Speech Steganography Using Differential SVD

et al. 2019

View full text Add to dashboard Cite

show abstract

“…However, with the popularization of smart devices, the application scenario of deep neural network applications has grown far beyond the high-performance platforms in their infancy. From computer vision (Redmon et al 2016) to image processing (Vardhana et al 2018), from audio analysis (Sehgal and Kehtarnavaz 2018) to natural language processing (Goldberg 2017), various edge portable and lowpower embedded platforms represented by smartphones have gradually become the main processing platforms for deep learning applications. The efficient and timely processing of deep learning applications on these embedded platforms has gradually become an increasingly important optimization design problem in deep learning research and practice.…”

Section: Neural Network Backgroundmentioning

confidence: 99%

To cloud or not to cloud: an on-line scheduler for dynamic privacy-protection of deep learning workload on edge devices

Tang

Wang

et al. 2020

CCF Trans. HPC

View full text Add to dashboard Cite

Recently deep learning applications are thriving on edge and mobile computing scenarios, due to the concerns of latency constraints, data security and privacy, and other considerations. However, because of the limitation of power delivery, battery lifetime and computation resource, offering real-time neural network inference ability has to resort to the specialized energy-efficient architecture, and sometimes the coordination between the edge devices and the powerful cloud or fog facilities. This work investigates a realistic scenario when an on-line scheduler is needed to meet the requirement of latency even when the edge computing resources and communication speed are dynamically fluctuating, while protecting the privacy of users as well. It also leverages the approximate computing feature of neural networks and actively trade-off excessive neural network propagation paths for latency guarantee even when local resource provision is unstable. Combining neural network approximation and dynamic scheduling, the real-time deep learning system could adapt to different requirements of latency/ accuracy and the resource fluctuation of mobile-cloud applications. Experimental results also demonstrate that the proposed scheduler significantly improves the energy efficiency of real-time neural networks on edge devices.

show abstract

“…Noting that multi-core processors are used in modern smartphones, a DNN model can be run on a secondary thread to create the needed computational bandwidth on the main thread to run the app at a desired FPS. This technique was used previously in [11] to allow a DNN model to run on a parallel thread by removing the computation burden from the main audio thread and thus preventing any audio frames from being skipped.…”

Section: Multithreadingmentioning

confidence: 99%

Guidelines and Benchmarks for Deployment of Deep Learning Models on Smartphones as Real-Time Apps

Sehgal

Kehtarnavaz

2019

MAKE

Self Cite

View full text Add to dashboard Cite

Deep learning solutions are being increasingly used in mobile applications. Although there are many open-source software tools for the development of deep learning solutions, there are no guidelines in one place in a unified manner for using these tools toward real-time deployment of these solutions on smartphones. From the variety of available deep learning tools, the most suited ones are used in this paper to enable real-time deployment of deep learning inference networks on smartphones. A uniform flow of implementation is devised for both Android and iOS smartphones. The advantage of using multi-threading to achieve or improve real-time throughputs is also showcased. A benchmarking framework consisting of accuracy, CPU/GPU consumption, and real-time throughput is considered for validation purposes. The developed deployment approach allows deep learning models to be turned into real-time smartphone apps with ease based on publicly available deep learning and smartphone software tools. This approach is applied to six popular or representative convolutional neural network models, and the validation results based on the benchmarking metrics are reported.

show abstract

A Convolutional Neural Network Smartphone App for Real-Time Voice Activity Detection

Cited by 99 publications

References 20 publications

Robust Speech Steganography Using Differential SVD

Robust Speech Steganography Using Differential SVD

To cloud or not to cloud: an on-line scheduler for dynamic privacy-protection of deep learning workload on edge devices

Guidelines and Benchmarks for Deployment of Deep Learning Models on Smartphones as Real-Time Apps

Contact Info

Product

Resources

About