Lightweight deep convolutional neural network for background sound classification in speech signals

Dayal, Aveen; Yeduri, Sreenivasa Reddy; Koduru, Balu Harshavardan; Jaiswal, Rahul; Soumya, J; Srinivas, M.B.; Pandey, Om Jee; Cenkeramaddi, Linga Reddy

doi:10.1121/10.0010257

Cited by 7 publications

(3 citation statements)

References 29 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…All of these will require detailed investigation and beyond the scope of this work. Additionally, in a real-world deployment, as the system make a transition from unguided to semi-guided or guided models with time pass, the system can identify different background sounds using sound classification approaches [60] , [61] , [62] and retrain the initial unguided cough model to obtain more robust models utilizing relevant background sounds by following approaches similar to the Federated learning, environment knowledge broadcast among users, and place discovery [63] , [64] , [65] , [66] , [67] , [68] . These are beyond the scope of this manuscript.…”

Section: Discussionmentioning

confidence: 99%

Environment Knowledge-Driven Generic Models to Detect Coughs From Audio Recordings

Vhaduri

Dibbo

Kim

2023

IEEE Open J. Eng. Med. Biol.

View full text Add to dashboard Cite

Goal: Millions of people are dying due to respiratory diseases, such as COVID-19 and asthma, which are often characterized by some common symptoms, including coughing. Therefore, objective reporting of cough symptoms utilizing environment-adaptive machine-learning models with microphone sensing can directly contribute to respiratory disease diagnosis and patient care. Methods: In this work, we present three generic modeling approaches – unguided , semi-guided , and guided approaches considering three potential scenarios, i.e., when a user has no prior knowledge, some knowledge, and detailed knowledge about the environments, respectively. Results: From detailed analysis with three datasets, we find that guided models are up to 28% more accurate than the unguided models. We find reasonable performance when assessing the applicability of our models using three additional datasets, including two open-sourced cough datasets. Conclusions: Though guided models outperform other models, they require a better understanding of the environment.

show abstract

Section: Discussionmentioning

confidence: 99%

Environment Knowledge-Driven Generic Models to Detect Coughs From Audio Recordings

Vhaduri

Dibbo

Kim

2023

IEEE Open J. Eng. Med. Biol.

View full text Add to dashboard Cite

show abstract

“…(1) the input is provided by raw audio data recorded by two hydrophones, which allows it to perform joint feature learning with passive whales, avoiding manual feature selection. Meanwhile, (2) an end-to-end data-driven approach brings us the possibility to capture more complex spatiotemporally correlated latent features of the two hydrophones through the main convolution operation (Chen and Schmidt, 2021;Dayal et al, 2022).…”

Section: Training Processmentioning

confidence: 99%

An acoustic tracking model based on deep learning using two hydrophones and its reverberation transfer hypothesis, applied to whale tracking

Jin,

Xu,

Zhang

et al. 2023

Front. Mar. Sci.

View full text Add to dashboard Cite

Acoustic tracking of whales’ underwater cruises is essential for protecting marine ecosystems. For cetacean conservationists, fewer hydrophones will provide more convenience in capturing high-mobility whale positions. Currently, it has been possible to use two hydrophones individually to accomplish direction finding or ranging. However, traditional methods only aim at estimating one of the spatial parameters and are susceptible to the detrimental effects of reverberation superimposition. To achieve complete whale tracking under reverberant interference, in this study, an intelligent acoustic tracking model (CIAT) is proposed, which allows both horizontal direction discrimination and distance/depth perception by mining unpredictable features of position information directly from the received signals of two hydrophones. Specifically, the horizontal direction is discriminated by an enhanced cross-spectral analysis to make full use of the exact frequency of received signals and eliminate the interference of non-source signals, and the distance/depth direction combines convolutional neural network (CNN) with transfer learning to address the adverse effects caused by unavoidable acoustic reflections and reverberation superposition. Experiments with real recordings show that 0.13 km/MAE is achieved within 8 km. Our work not only provides satisfactory prediction performance, but also effectively avoids the reverberation effect of long-distance signal propagation, opening up a new avenue for underwater target tracking.

show abstract

“…CNNs also show great potential when being used on spectrogram data based on time-frequency representations of spatiotemporal signals. This has been applied to audio signals (Dayal et al, 2022), speech recognition (Badshah et al, 2017), but even in the context of gait classification (Jung et al, 2019). In clinical contexts with PD, recent studies have used this approach on EEG data (Khare et al, 2021;.…”

Section: Introductionmentioning

confidence: 99%

Convolutional neural network reveals frequency content of medio-lateral COM body sway to be highly predictive of Parkinson’s disease

Engel

Greulich

Parola

et al. 2023

Preprint

View full text Add to dashboard Cite

Postural instability as a symptom of progressing Parkinson's disease (PD) greatly reduces quality of life. Hence, early detection of postural impairments is crucial to facilitate interventions. Our aim was to use a convolutional neural network (CNN) to differentiate people with early to mid-stage PD from healthy age-matched individuals based on spectrogram images obtained from their body movement. We hypothesized the time-frequency content of body sway to be predictive of PD, even when impairments are not yet manifested in day-to-day postural control. We tracked their center of pressure (COP) using a Wii Balance Board and their full-body motion using a Microsoft Kinect, out of which we calculated the trajectory of their center of mass (COM). We used 30 s-snippets of motion data from which we acquired wavelet-based time-frequency spectrograms that were fed into a custom-built CNN as labeled images. We used binary classification to have the network differentiate between individuals with PD and controls (n=15, respectively). Classification performance was best when the medio-lateral motion of the COM was considered. Here, our network reached an average predictive accuracy of 98.45 % with a receiver operating characteristic area under the curve of 1.0. Moreover, an explainable AI approach revealed high frequencies in the postural sway data to be most distinct between both groups. Our findings suggest a CNN classifier based on cost-effective and conveniently obtainable posturographic data to be a promising approach to detect postural impairments in early to mid-stage PD and to gain novel insight into the subtle characteristics of impairments at this stage of the disease.

show abstract

Lightweight deep convolutional neural network for background sound classification in speech signals

Cited by 7 publications

References 29 publications

Environment Knowledge-Driven Generic Models to Detect Coughs From Audio Recordings

Environment Knowledge-Driven Generic Models to Detect Coughs From Audio Recordings

An acoustic tracking model based on deep learning using two hydrophones and its reverberation transfer hypothesis, applied to whale tracking

Convolutional neural network reveals frequency content of medio-lateral COM body sway to be highly predictive of Parkinson’s disease

Contact Info

Product

Resources

About