Metrics for Polyphonic Sound Event Detection

Mesaros, Annamaria; Heittola, Toni; Virtanen, Tuomas

doi:10.3390/app6060162

Cited by 463 publications

(330 citation statements)

References 33 publications

Supporting

Mentioning

309

Contrasting

Unclassified

Order By: Relevance

“…Another evaluation of the proposed LDS-based event tracking method is made when comparing LDS with 32 provide a small but consistent performance improvement over that which considers only as latent variables the individual event activations per sound class.…”

Section: E Resultsmentioning

confidence: 99%

“…For evaluation, we employed event detection metrics both from the DCASE 2013 challenge [1], as well as the upcoming DCASE 2016 challenge [32]. Specifically, 3 different metrics are used: frame-based (used in DCASE 2013), segment-based (used in DCASE 2016), and class-wise segment-based (used in DCASE 2016).…”

Section: Metricsmentioning

confidence: 99%

“…Finally, class-wise segment-based metrics also consider 100 msec segment size, with the results being normalized per class. A key difference between the frame-based and segmentbased metrics is that frame-based metrics are computed per recording and are averaged across the entire dataset, whereas segment-based metrics count the number of true positives, false positives and false negatives across the entire dataset prior to the metrics computation [32].…”

Section: Metricsmentioning

confidence: 99%

See 2 more Smart Citations

Polyphonic Sound Event Tracking Using Linear Dynamical Systems

Benetos

Lafay

Lagrange

et al. 2017

IEEE/ACM Trans. Audio Speech Lang. Process.

View full text Add to dashboard Cite

Abstract-In this paper, a system for polyphonic sound event detection and tracking is proposed, based on spectrogram factorisation techniques and state space models. The system extends probabilistic latent component analysis (PLCA) and is modelled around a 4-dimensional spectral template dictionary of frequency, sound event class, exemplar index, and sound state. In order to jointly track multiple overlapping sound events over time, the integration of linear dynamical systems (LDS) within the PLCA inference is proposed. The system assumes that the PLCA sound event activation is the (noisy) observation in an LDS, with the latent states corresponding to the true event activations. LDS training is achieved using fully observed data, making use of ground truth-informed event activations produced by the PLCA-based model. Several LDS variants are evaluated, using polyphonic datasets of office sounds generated from an acoustic scene simulator, as well as real and synthesized monophonic datasets for comparative purposes. Results show that the integration of LDS tracking within PLCA leads to an improvement of +8.5-10.5% in terms of frame-based F-measure as compared to the use of the PLCA model alone. In addition, the proposed system outperforms several state-of-the-art methods for the task of polyphonic sound event detection.

show abstract

Section: E Resultsmentioning

confidence: 99%

Section: Metricsmentioning

confidence: 99%

Section: Metricsmentioning

confidence: 99%

See 1 more Smart Citation

Polyphonic Sound Event Tracking Using Linear Dynamical Systems

Benetos

Lafay

Lagrange

et al. 2017

IEEE/ACM Trans. Audio Speech Lang. Process.

View full text Add to dashboard Cite

show abstract

“…In the study of Mesaros et al, various metrics recommended for assessment of polyphonic sound event perception systems used in realist cases, where multiple sound sources are simultaneously active, are presented and discussed [16].…”

Section: Related Workmentioning

confidence: 99%

Wearable Vibration Based Computer Interaction and Communication System for Deaf

Yağanoğlu

Köse

2017

Applied Sciences

View full text Add to dashboard Cite

Abstract:In individuals with impaired hearing, determining the direction of sound is a significant problem. The direction of sound was determined in this study, which allowed hearing impaired individuals to perceive where sounds originated. This study also determined whether something was being spoken loudly near the hearing impaired individual. In this manner, it was intended that they should be able to recognize panic conditions more quickly. The developed wearable system has four microphone inlets, two vibration motor outlets, and four Light Emitting Diode (LED) outlets. The vibration of motors placed on the right and left fingertips permits the indication of the direction of sound through specific vibration frequencies. This study applies the ReliefF feature selection method to evaluate every feature in comparison to other features and determine which features are more effective in the classification phase. This study primarily selects the best feature extraction and classification methods. Then, the prototype device has been tested using these selected methods on themselves. ReliefF feature selection methods are used in the studies; the success of K nearest neighborhood (Knn) classification had a 93% success rate and classification with Support Vector Machine (SVM) had a 94% success rate. At close range, SVM and two of the best feature methods were used and returned a 98% success rate. When testing our wearable devices on users in real time, we used a classification technique to detect the direction and our wearable devices responded in 0.68 s; this saves power in comparison to traditional direction detection methods. Meanwhile, if there was an echo in an indoor environment, the success rate increased; the echo canceller was disabled in environments without an echo to save power. We also compared our system with the localization algorithm based on the microphone array; the wearable device that we developed had a high success rate and it produced faster results at lower cost than other methods. This study provides a new idea for the benefit of deaf individuals that is preferable to a computer environment.

show abstract

“…In this work, segment-based F-score in a fixed time grid [26] is adopted as the evaluation metric. A segment-based metric is performed in short segments.…”

Section: Evaluation Metricmentioning

confidence: 99%

Frame-Wise Dynamic Threshold Based Polyphonic Acoustic Event Detection

et al. 2017

View full text Add to dashboard Cite

Acoustic event detection, the determination of the acoustic event type and the localisation of the event, has been widely applied in many real-world applications. Many works adopt multi-label classification techniques to perform the polyphonic acoustic event detection with a global threshold to detect the active acoustic events. However, the global threshold has to be set manually and is highly dependent on the database being tested. To deal with this, we replaced the fixed threshold method with a frame-wise dynamic threshold approach in this paper. Two novel approaches, namely contour and regressor based dynamic threshold approaches are proposed in this work. Experimental results on the popular TUT Acoustic Scenes 2016 database of polyphonic events demonstrated the superior performance of the proposed approaches.

show abstract

Metrics for Polyphonic Sound Event Detection

Cited by 463 publications

References 33 publications

Polyphonic Sound Event Tracking Using Linear Dynamical Systems

Polyphonic Sound Event Tracking Using Linear Dynamical Systems

Wearable Vibration Based Computer Interaction and Communication System for Deaf

Frame-Wise Dynamic Threshold Based Polyphonic Acoustic Event Detection

Contact Info

Product

Resources

About