Multi-speaker activity detection using zero crossing rate

Ramaiah, V. Subba; Rao, R. Srinivasa

doi:10.1109/iccsp.2016.7754232

Cited by 14 publications

(3 citation statements)

References 10 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Soundscape analysis was performed using eco-acoustic indices in R v.3.5.3 [9] with the "soundecology" package [10]. The eco-acoustic indices employed in this work were calculated using an FFT 2048 point and with the following frequency values: Acoustic Complexity Index (ACI) [11] (freq.range=500-12.000 Hz) Acoustic Diversity Index (ADI) [12] (freq.max = 12.000 Hz, freq.step = 10 Hz) Acoustic Evenness Index (AEI) [12] (freq.max = 12.000 Hz, freq.step= 10 Hz) Bio-acoustic Index (BI) [13] (freq.range=2.000-12.000 Hz) Acoustic Entropy Index (H) [14] (freq.range= 0-20.000Hz) Normalized Difference Soundscape Index (NDSI) [15] (anthropic.freq.range= 500-1500Hz, bio.freq.range= 1500-12000 Hz); Dynamic Spectral Centroid (DSC) [12] (freq.range = 500-12.000 Hz) Zero-Crossing Rate (ZCR) [16,17] (freq.range= 500-12.000Hz)…”

Section: Calculation Of Eco-acoustic Indicesmentioning

confidence: 99%

Soundscape of an Urban Green Area of Milan

Zaffaroni-Caorsi,

Benocci,

Potenza

et al. 2024

Proceedings of the 10th Convention of the European Acoustics Association Forum Acusticum 2023

View full text Add to dashboard Cite

Over the last century, anthropogenic activities have caused significant changes in soundscapes. These alterations are mainly caused by growing noise related to the expansion of cities and their connections (e.g. roads). It is urgent to understand the distribution of noise pollution and its effect on biodiversity, especially in large cities, aiming to propose legal regulations and mitigation actions to reduce its impact. In this study, we assessed the soundscape of the pocketpark Vivaio-Bicocca of Milan (Italy), in a highly anthropized urban context, as a starting point to understand the effect of noise pollution on the city wildlife. Audio recordings took place 24h a day from May 31st to June 14th 2022 (1 minute of recordings followed by 5-minutes pause) with autonomous recorders in three sites 45m apart, while noise level was measured using a Sound-Level-Meter. Soundscape analyses were performed calculating ecoacoustic indices (e.g. Acoustic Complexity Index, Acoustic Diversity Index, Normalized Difference Soundscape Index) in the R environment. As expected, the soundscape of the three sites resulted quite similar, composed of biophony and a high quantity of anthropophony. We hypothesize that the central site could be representative of the whole park and be used on its own for the soundscape characterization.

show abstract

Section: Calculation Of Eco-acoustic Indicesmentioning

confidence: 99%

Soundscape of an Urban Green Area of Milan

Zaffaroni-Caorsi,

Benocci,

Potenza

et al. 2024

Proceedings of the 10th Convention of the European Acoustics Association Forum Acusticum 2023

View full text Add to dashboard Cite

show abstract

“…(2) implementation of the eco-acoustic indices (ACI, ADI, AEI, BI, NDSI, H, DSC, ZCR) using an FFT of 1024 points [7,8]; computation of the percentage error on eco-acoustic indices (3).…”

Section: Equalization Procedures and Data Analysismentioning

confidence: 99%

Audio Equalization of Different Soundscape Recorders is Necessary to Perform a Correct Analysis

Potenza,

Benocci,

Bisceglie

et al. 2022

Proceedings of the 10th Convention of the European Acoustics Association Forum Acusticum 2023

View full text Add to dashboard Cite

Passive acoustic monitoring (PAM) is widely used by soundscape and bioacoustics researchers. Nowadays, there is a large number of instruments available to record environmental sound, with different characteristics and costs. Given this availability, using all the devices on hand, even from different manufacturers, can be tempting since it allows to expand the area of study with less costs. Unfortunately, due to their diverse frequency response, this procedure introduces biases in a comparative analysis of the recordings (e.g., the computation of acoustic parameters and indices, and the following elaborations). For these reasons, equalization of audio recording is fundamental to ensure proper analysis. Therefore, in this study, we evaluated the effects of the equalization on the data acquired by two soundscape recorder typologies (Song-Meter-Micro and Soundscape-Explorer-Terrestrial). The evaluation was carried out by computing the eco-acoustic indices. The equalization was performed in a laboratory: the soundscape recorders and a class 1 sound level meter (able to produce a .WAV file) were exposed to a white noise source. The equalization curve was calculated in MATLAB environment with a successful response. As a second step, its effect on the eco-acoustic indices was evaluated by applying it to audio recordings from a park in Milan (Italy).

show abstract

“…Some researchers believe that features based on feature engineering are still more efficient than end-to-end models based on feature learning in some cases [ 15 ]. Typical examples of extraction from the time domain are zero crossing rate [ 16 ], autocorrelation coefficients [ 17 ], and the time-domain envelope [ 18 ]. Compared with the time-domain features, there are more optional features in the frequency domain, such as spectral envelope [ 19 ], spectral flux [ 20 ], and Mel spectrogram [ 21 ].…”

Section: Introductionmentioning

confidence: 99%

A Parallel Classification Model for Marine Mammal Sounds Based on Multi-Dimensional Feature Extraction and Data Augmentation

Cai

Zhu

Zhang

et al. 2022

Sensors

View full text Add to dashboard Cite

Due to the poor visibility of the deep-sea environment, acoustic signals are often collected and analyzed to explore the behavior of marine species. With the progress of underwater signal-acquisition technology, the amount of acoustic data obtained from the ocean has exceeded the limit that human can process manually, so designing efficient marine-mammal classification algorithms has become a research hotspot. In this paper, we design a classification model based on a multi-channel parallel structure, which can process multi-dimensional acoustic features extracted from audio samples, and fuse the prediction results of different channels through a trainable full connection layer. It uses transfer learning to obtain faster convergence speed, and introduces data augmentation to improve the classification accuracy. The k-fold cross-validation method was used to segment the data set to comprehensively evaluate the prediction accuracy and robustness of the model. The evaluation results showed that the model can achieve a mean accuracy of 95.21% while maintaining a standard deviation of 0.65%. There was excellent consistency in performance over multiple tests.

show abstract

Multi-speaker activity detection using zero crossing rate

Cited by 14 publications

References 10 publications

Soundscape of an Urban Green Area of Milan

Soundscape of an Urban Green Area of Milan

Audio Equalization of Different Soundscape Recorders is Necessary to Perform a Correct Analysis

A Parallel Classification Model for Marine Mammal Sounds Based on Multi-Dimensional Feature Extraction and Data Augmentation

Contact Info

Product

Resources

About