Real Time Blind Audio Source Separation Based on Machine Learning Algorithms

Alghamdi, Arwa; Healy, Graham; Abdelhafez, Hoda Ahmed

doi:10.1109/niles50944.2020.9257891

Cited by 3 publications

(5 citation statements)

References 12 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…AI-based BSS algorithms utilize neural networks, i.e., classical or shallow models, evolutionary algorithms, and deep learning architectures to learn the optimal separation coefficients and minimize the statistical dependency among the estimated components [27], [59], [60], [61], [62], [63], [64], [65], [66]. The utilization of neural networks in the BSS methodology plays a crucial role in reducing the statistical dependency among the estimated components.…”

Section: B Artificial Intelligence (Ai)-based Methodsmentioning

confidence: 99%

“…The study of [65] highlights the capability of machine learning algorithms, specifically convolutional time-domain audio separation network (Conv-TasNet) and deep extractor for music sources (Demucs), to discriminate between two interfering signals (such as speech and music) without prior knowledge of the mixture operation. The Demucs algorithm is a waveform-to-waveform model that exhibits a higher decoding capacity compared to the Conv-TasNet model, leveraging the same technique as the audio generation algorithm.…”

Section: B Artificial Intelligence (Ai)-based Methodsmentioning

confidence: 99%

See 1 more Smart Citation

A Robust Hybrid Neural Network Architecture for Blind Source Separation of Speech Signals Exploiting Deep Learning

Ansari,

Alnajjar,

Khater

et al. 2023

IEEE Access

View full text Add to dashboard Cite

In the contemporary era, blind source separation has emerged as a highly appealing and significant research topic within the field of signal processing. The imperative for the integration of blind source separation techniques within the context of beyond fifth-generation and sixth-generation networks arises from the increasing demand for reliable and efficient communication systems that can effectively handle the challenges posed by high-density networks, dynamic interference environments, and the coexistence of diverse signal sources, thereby enabling enhanced signal extraction and separation for improved system performance. Particularly, audio processing presents a critical domain where the challenge lies in effectively handling files containing a mixture of human speech, silence, and music. Addressing this challenge, speech separation systems can be regarded as a specialized form of human speech recognition or audio signal classification systems that are leveraged to separate, identify, or delineate segments of audio signals encompassing human speech. In various applications such as volume reduction, quality enhancement, detection, and identification, the need arises to separate human speech by eliminating silence, music, or environmental noise from the audio signals. Consequently, the development of robust methods for accurate and efficient speech separation holds paramount importance in optimizing audio signal processing tasks. This study proposes a novel three-way neural network architecture that incorporates transfer learning, a pre-trained dual-path recurrent neural network, and a transformer. In addition to learning the time series associated with audio signals, this network possesses the unique capability of direct context-awareness for modeling the speech sequence within the transformer framework. A comprehensive array of simulations is meticulously conducted to evaluate the performance of the proposed model, which is benchmarked with seven prominent state-of-the-art deep learning-based architectures. The results obtained from these evaluations demonstrate notable advancements in multiple objective metrics. Specifically, our proposed solution showcases an average improvement of 4.60% in terms of short-time objective intelligibility, 14.84% in source-to-distortion ratio, and 9.87% in scale-invariant signal-to-noise ratio. These extraordinary advancements surpass those achieved by the nearest rival, namely the dual-path recurrent neural network time-domain audio separation network, firmly establishing the superiority of our proposed model's performance.

show abstract

Section: B Artificial Intelligence (Ai)-based Methodsmentioning

confidence: 99%

Section: B Artificial Intelligence (Ai)-based Methodsmentioning

confidence: 99%

A Robust Hybrid Neural Network Architecture for Blind Source Separation of Speech Signals Exploiting Deep Learning

Ansari,

Alnajjar,

Khater

et al. 2023

IEEE Access

View full text Add to dashboard Cite

show abstract

“…Also, discuss the evaluation method via the performance of the algorithm and the input and output signals. This research study was extended [1] to apply the Natural Language (NL) of three different languages English, Arabic and Chinese.…”

Section: Methodsmentioning

confidence: 99%

“…In [20], the authors speed up the Conv-TasNet during source separation to enhance the execution time which supports a real-time methodology [21]. This paper is an extension of work [1] to study the effectiveness of the Natural Language on the separation process of Demucs and Conv-TasNet algorithms. The experiments' inputs were English, Arabic and Chinese Audioset.…”

Section: Related Workmentioning

confidence: 99%

See 1 more Smart Citation

Machine Learning Algorithms for Real Time Blind Audio Source Separation with Natural Language Detection

Alghamdi¹,

Healy²,

Abdelhafez³

2021

Adv. sci. technol. eng. syst. j.

Self Cite

View full text Add to dashboard Cite

The Conv-TasNet and Demucs algorithms, can differentiate between two mixed signals, such as music and speech, the mixing operation proceed without any support information. The network of convolutional time-domain audio separations is used in Conv-TasNet algorithm, while there is a new waveform-to-waveform model in Demucs algorithm. The Demucs algorithm utilizes a procedure like the audio generation model and sizable decoder capacity. The algorithms are not pretrained; so, the process of separation is blindly without any function of three Natural Languages (NL) detection. This research study evaluated the quality and execution time of the separation output signals. It focused on studying the effectiveness of NL in Both algorithms based on four sound signal experiments: (music & male), (music &female), (music & conversation), and finally (music & child). In addition, this research studies three NL, which are English, Arabic and Chinese. The results are evaluated based on R square and mir_eval libraries, mean absolute Error (MAE) scores and root mean square error (RMSE). Conv-TasNet has the highest Signal-to-distortion-Ratio (SDR) score is 9.21 of music at (music & female) experiment, and the highest SDR value of child signal is 8.14. The SDR score of music at (music & female) experiment is 7.8 during the Demucs algorithm, whereas child output signal has the highest SDR score 8.15. However, the average execution time of English experiment of Conv-TasNet is seven times faster than Demucs. For accuracy measure, RMSE indicates absolute values, and MAE handles the errors between observations and prediction signals. Both algorithms show high accuracy and excellent results in the separation process.

show abstract

Background Music Removal Using Deep Learning

Ozawa,

Ogawa

2023

2023 IEEE 3rd International Conference on Software Engineering and Artificial Intelligence (SEAI)

View full text Add to dashboard Cite

Real Time Blind Audio Source Separation Based on Machine Learning Algorithms

Cited by 3 publications

References 12 publications

A Robust Hybrid Neural Network Architecture for Blind Source Separation of Speech Signals Exploiting Deep Learning

A Robust Hybrid Neural Network Architecture for Blind Source Separation of Speech Signals Exploiting Deep Learning

Machine Learning Algorithms for Real Time Blind Audio Source Separation with Natural Language Detection

Background Music Removal Using Deep Learning

Contact Info

Product

Resources

About