Use of local kurtosis measure for spotting usable speech segments in co-channel speech

Krishnamachari, K.R.; Yantorno, Robert E.; Lovekin, J.M.; Benincasa, D.S.; Wenndt, Stanley J.

doi:10.1109/icassp.2001.940915

Cited by 15 publications

(6 citation statements)

References 11 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Additionally, we found that features which were originally designed for a different purpose can also play a role in crosstalk analysis (e.g., "fundamentalness" [14]). Furthermore, features which have previously been used to identify overlapping speakers such as MFCCs, PPF [16], and SAPVR (e.g., [8]) were rejected.…”

Section: Discussionmentioning

confidence: 99%

“…They demonstrated that the kurtosis of overlapped speech is generally less than the kurtosis of isolated utterances, since-in accordance with the central limit theorem-mixtures of speech signals will tend toward a Gaussian distribution. This statistical property has also been used to identify reliable frames for speaker identification in the presence of an interfering talker [8].…”

mentioning

confidence: 99%

See 1 more Smart Citation

Speech and crosstalk detection in multichannel audio

Wrigley

Brown

Wan

et al. 2005

IEEE Trans. Speech Audio Process.

View full text Add to dashboard Cite

Abstract-The analysis of scenarios in which a number of microphones record the activity of speakers, such as in a round-table meeting, presents a number of computational challenges. For example, if each participant wears a microphone, speech from both the microphone's wearer (local speech) and from other participants (crosstalk) is received. The recorded audio can be broadly classified in four ways: local speech, crosstalk plus local speech, crosstalk alone and silence. We describe two experiments related to the automatic classification of audio into these four classes. The first experiment attempted to optimize a set of acoustic features for use with a Gaussian mixture model (GMM) classifier. A large set of potential acoustic features were considered, some of which have been employed in previous studies. The best-performing features were found to be kurtosis, "fundamentalness," and cross-correlation metrics. The second experiment used these features to train an ergodic hidden Markov model classifier. Tests performed on a large corpus of recorded meetings show classification accuracies of up to 96%, and automatic speech recognition performance close to that obtained using ground truth segmentation.

show abstract

Section: Discussionmentioning

confidence: 99%

mentioning

confidence: 99%

Speech and crosstalk detection in multichannel audio

Wrigley

Brown

Wan

et al. 2005

IEEE Trans. Speech Audio Process.

View full text Add to dashboard Cite

show abstract

“…Kurtosis has been reported as an effective measure for detecting the presence of multiple-speakers in co-channel signals in numerous studies [5,4,19]. It is well-known that overlapped-speech signals exhibit lower kurtosis compared to single-speaker speech signals [20].…”

Section: Kurtosismentioning

confidence: 99%

Overlapped-speech detection with applications to driver assessment for in-vehicle active safety systems

Shokouhi

Sathyanarayana

Sadjadi

et al. 2013

2013 IEEE International Conference on Acoustics, Speech and Signal Processing

View full text Add to dashboard Cite

In this study we propose a system for overlapped-speech detection. Spectral harmonicity and envelope features are extracted to represent overlapped and single-speaker speech using Gaussian mixture models (GMM). The system is shown to effectively discriminate the single and overlapped speech classes. We further increase the discrimination by proposing a phoneme selection scheme to generate more reliable artificial overlapped data for model training. Evaluations on artificially generated co-channel data show that the novelty in feature selection and phoneme omission results in a relative improvement of 10% in the detection accuracy compared to baseline. As an example application, we evaluate the effectiveness of overlapped-speech detection for vehicular environments and its potential in assessing driver alertness. Results indicate a good correlation between driver performance and the amount and location of overlapped-speech segments.

show abstract

“…• Spectral autocorrelation peak-to-valley ratio (SAPVR) [39,40] • Local kurtosis [41] • Adjacent pitch period comparison (APPC) [42] • Cyclostationarity and wavelet transform [43] • Linear predictive analysis [44,45] • Difference-mean comparison (DMC) and nodal density (ND) [46] To ensure that an efficient usable speech detection system was developed, several usable speech detection features containing complementary information were fused [38,47]. As a result, the overall performance of the detection system was enhanced.…”

Section: Previous Work 25mentioning

confidence: 99%

Co-channel speech separation using state-space reconstruction and sinusoidal modelling

Mahgoub¹

View full text Add to dashboard Cite

Use of local kurtosis measure for spotting usable speech segments in co-channel speech

Cited by 15 publications

References 11 publications

Speech and crosstalk detection in multichannel audio

Speech and crosstalk detection in multichannel audio

Overlapped-speech detection with applications to driver assessment for in-vehicle active safety systems

Co-channel speech separation using state-space reconstruction and sinusoidal modelling

Contact Info

Product

Resources

About