MétaCan
Menu
Back to cohort
Record W7026848749

Active noise cancelation on construction sites using advanced deep learning

2023· dissertation· en· W7026848749 on OpenAlex

Why this work is in the frame

A frame that forgets how it found something cannot be audited. These are the routes that admitted this work.

affAt least one author lists a Canadian institution in the pinned OpenAlex snapshot.

Bibliographic record

VenueMspace (University of Manitoba) · 2023
Typedissertation
Languageen
FieldBiochemistry, Genetics and Molecular Biology
TopicChemical and Physical Studies
Canadian institutionsUniversity of Manitoba
Fundersnot available
KeywordsActive noise controlFeed forwardNoise reductionScalabilityNonlinear systemDeep learningNoise (video)Rendering (computer graphics)
DOInot available

Abstract

fetched live from OpenAlex

This thesis proposes novel active noise cancellation (ANC) algorithms based on deep learning to mitigate non-stationary and nonlinear noise. Traditional active noise control (ANC) methods face challenges when it comes to effectively mitigating construction-related noise, primarily due to the nonlinearity and transient nature of machinery sounds encountered on construction sites. In order to address this limitation, a highly effective feedforward ANC controller, named construction site noise network (CsNNet), has been developed utilizing advanced deep learning techniques. The proposed algorithm incorporates considerations for acoustic device delay and nonlinear characteristics, rendering it particularly suitable for open environments such as construction sites. Through extensive simulation studies, CsNNet demonstrated remarkable broadband noise reduction capabilities, achieving an average attenuation of approximately 8.3 dB across a wide range of construction-related noises. These results surpass the performance of both traditional ANC algorithms and contemporary state-of-the-art approaches. Additionally, CsNNet offers the advantage of scalability to multichannel ANC control without incurring additional computational costs, distinguishing it from previously developed ANC algorithms. Following extensive simulations and the development of the network architecture, we proceeded to assess the algorithm's performance through experimental testing in an acoustic environment. We carefully selected suitable equipment for the ANC system, including the microphone, loudspeaker, and signal acquisition device, prioritizing quality and minimizing delays. To accurately capture the characteristics of real-life acoustic and electrical secondary paths, we introduced a novel secondary path model based on deep learning. This model effectively addressed the limitations of traditional methods that relied on linear finite impulse response (FIR) filters for secondary path modeling. By incorporating this precise secondary path model, we conducted experimental tests on the causal version of CsNNet and observed a consistent agreement between the simulation and experimental results. The proposed algorithm is a significant contribution to the field of ANC using deep learning. It can be applied to various environments and has practical implications for noise control in different fields. The algorithm shows superior performance in controlling construction-related noise, which is a severe issue for governments in metropolitan cities. It has the potential to improve the quality of life in urban environments and reduce the impact of noise pollution on human health. The algorithm can also be used for noise control in other fields like transportation and aviation, where noise pollution is a significant issue. Overall, the thesis presents significant contributions to the field of ANC using deep learning-based algorithms, which have the potential to revolutionize noise control techniques. It is important to highlight that the content of this thesis is derived from our paper [1], with additional explanations provided for each section and an experimental investigation of the algorithm, along with the corresponding results.

Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.

Full frame distilled prediction

Teacher imitation

Not calibrated prevalence, not ground truth. Human validation pending. Learned from the 10,348 direct Codex labels and 10,348 direct Gemma labels. Candidate is the union of thresholded teacher heads; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels or direct frontier model labels.

metaresearch head score (Codex)0.000
metaresearch head score (Gemma)0.000
Version: codex-gemma-dda1882f352aValidation status: machine_predicted_unvalidated
Candidate categoriesnone
Consensus categoriesnone
DomainCandidate signal: none · Consensus signal: none
Study designCandidate signal: Bench or experimental · Consensus signal: Bench or experimental
GenreCandidate signal: Empirical · Consensus signal: Empirical
Teacher disagreement score0.418
Threshold uncertainty score0.986

Codex and Gemma teacher scores by category

CategoryCodexGemma
Metaresearch0.0000.000
Meta-epidemiology (narrow)0.0000.000
Meta-epidemiology (broad)0.0000.000
Bibliometrics0.0000.000
Science and technology studies0.0000.000
Scholarly communication0.0000.000
Open science0.0000.000
Research integrity0.0000.000
Insufficient payload (model declined to judge)0.0000.000

Machine scores (provisional)

The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.

Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.

Opus teacher head0.011
GPT teacher head0.222
Teacher spread0.211 · how far apart the two teachers sit on this one work
Validation statusscore_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from it