A novel approach to robust speech endpoint detection in car environments

Huang, Liangsheng; Yang, Chung-Ho

doi:10.1109/icassp.2000.862091

Cited by 42 publications

(7 citation statements)

References 4 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Various studies have been conducted on the endpoint detection regarding speech signals. (41)(42)(43) The main task of endpoint detection of speech signals is to detect the start and end points of a speech signal. The collected acceleration data in Figs.…”

Section: Feature Extractionmentioning

confidence: 99%

“…Among various endpoint detection techniques, energy-based methods are the most widely applied solutions to this problem. (41) These algorithms are mostly based on the short-time Fourier transform (STFT). However, STFT has a fixed resolution, which might lead to poor time/frequency resolution in the analysis of nonstationary signals, e.g., a speech signal and a time series of acceleration collected here are both nonstationary signals.…”

Section: Feature Extractionmentioning

confidence: 99%

See 1 more Smart Citation

Automatic Labeling Framework for Wearable Sensor-based Human Activity Recognition

Liang¹,

Qing-sheng²,

Jia³

2018

Sensors and Materials

View full text Add to dashboard Cite

Labeled datasets are one of the key factors for obtaining a good and robust classifier using supervised learning methods. However, labeling raw data is a tedious and labor-intensive process, which is usually done manually. Many efforts were proposed to utilize a small amount of labeled data to train a classifier that is sufficiently robust to label more data for training or make a prediction on unlabeled data. Unlike previous studies, we proposed an automatic labeling framework without labeling a small amount of data in advance, to directly annotate unlabeled time series data regarding body-worn sensor-based human activity recognition (HAR) in laboratory settings. The framework automatically labels collected time series activity data by transforming the original data into its corresponding absolute wavelet energy entropy and detects activity endpoints based on constraints and information extracted from a predefined human activity sequence. The performance of the proposed framework was evaluated on the collected dataset and the UCI HAR Dataset. In both cases, the average precision and recall scores are above 81.9% and the average F-measure scores are above 88.9%. Results showed that the proposed framework can be adopted as a rapid and reliable way of generating labeled datasets from unlabeled data.

show abstract

Section: Feature Extractionmentioning

confidence: 99%

Section: Feature Extractionmentioning

confidence: 99%

Automatic Labeling Framework for Wearable Sensor-based Human Activity Recognition

Liang¹,

Qing-sheng²,

Jia³

2018

Sensors and Materials

View full text Add to dashboard Cite

show abstract

“…(2) Frequency-Domain detection method, such as LPC Cepstrum [4], Spectrum Entropy [5], Energy spectrum entropy [6][7], Distance Entropy [8], Sub-band Energy [9], Harmonic Energy [10], etc. (3)Nonlinear feature detection method, such as Permutation Entropy [11], C0 measure [12], HHT [13] , etc.…”

Section: Introductionmentioning

confidence: 99%

An Adaptive Voice Activity Detection Algorithm

Zhang

Huang

2015

International Journal on Smart Sensing and Intelligent Systems

View full text Add to dashboard Cite

Voice Activity Detection (VAD) is a crucial step for speech processing, which detecting accuracy and speed directly affects the effect of subsequent processing. Some voice processing system based phone or in the indoor environment, which need simple and quick method of VAD, for these representative voice signal, this paper proposes a new algorithm which is adaptive and quick based on a major improvement to Dual-Threshold endpoint detection algorithm. First the amplitude normalization is processed to the original voice signal, the characteristic is extracted by means of short-time amplitude, which can simplify operation. Then, large-scale (long frame-length and frame-shift) short-time amplitude is used for rough detection, combining adaptive threshold judgement of consecutive frames, which can find voice areas of start-point and end-point quickly. To these areas, small-scale (short frame-length and frame-shift) short-time amplitude is used for accurate detection, forward scanning is put to start-point area, reverse scanning is put to end-point area, combining adaptive threshold judgement of consecutive frames, start-point and end-point of the effective speech can be accurately located. Experimental results show that the method of this paper can detect endpoints of voice signal more quickly and accurately, which can improve recognition performance dramatically. Largescale can increase detection speed, small-scale can improve detection accuracy, both can be adjusted to satisfy the different requirements. The method of this paper ensures both detection speed and precision, which has more flexibility and applicability.

show abstract

“…More recently, there has been some interest in robustness to noise, due to the advent of cellular phones and handsfree headsets. For instance, there is the work of Junqua et al [1] which presents a number of adaptive energy-based techniques, the work of Huang and Yang [2], which uses This work was done while the author was at the MIT Media Laboratory a spectral entropy measure to pick out voiced regions, and later the work of Wu and Lin [3], which extends the work of Junqua et. al by looking at multiple bands and using a neural network to learn the appropriate thresholds.…”

Section: Introductionmentioning

confidence: 99%

“…We thus expect the entropy of a distribution taking this form to be relatively high. This leads us the notion of spectral entropy, as introduced by Huang and Yang [2].…”

Section: Spectral Entropymentioning

confidence: 99%

A linked-HMM model for robust voicing and speech detection

Basu

2003 IEEE International Conference on Acoustics, Speech, and Signal Processing, 2003. Proceedings. (ICASSP '03).

View full text Add to dashboard Cite

We present a novel method for simultaneous voicing and speech detection based on a linked-HMM architecture, with robust features that are independent of the signal energy. Because this approach models the change in dynamics between speech and non-speech regions, it is robust to low sampling rates, significant levels of additive noise, and large distances from the microphone. We demonstrate the performance of our method in a variety of testing conditions and also compare it to other methods reported in the literature.

show abstract

A novel approach to robust speech endpoint detection in car environments

Cited by 42 publications

References 4 publications

Automatic Labeling Framework for Wearable Sensor-based Human Activity Recognition

Automatic Labeling Framework for Wearable Sensor-based Human Activity Recognition

An Adaptive Voice Activity Detection Algorithm

A linked-HMM model for robust voicing and speech detection

Contact Info

Product

Resources

About