Showing papers by "Yaniv Zigel published in 2008"
14 Oct 2008
TL;DR: A unique and inexpensive fall detection system that does not require subjects to wear anything based on floor vibration and acoustic sensing, and uses a pattern recognition algorithm to discriminate between human or inanimate object fall events.
Abstract: Falls are very prevalent among the elderly especially in their home. The statistics show that approximately one in every three adults 65 years old or older falls each year. Almost 30% of those falls result in serious injuries. Studies have shown that the medical outcome of a fall is largely dependent upon the response and rescue time. Therefore, reliable and immediate fall detection system is important so that adequate medical support could be delivered. We have developed a unique and inexpensive solution that does not require subjects to wear anything. The solution is based on floor vibration and acoustic sensing, and uses a pattern recognition algorithm to discriminate between human or inanimate object fall events. Using the proposed system we can detect human falls with a sensitivity of 95% and specificity of 95%.
91 citations
12 May 2008
TL;DR: Experimental investigation is performed, using simulated and measured room impulse responses, NIST-based speech database, and AGMM based speaker verification system, showing significant improvement in performance.
Abstract: Speech recorded by a distant microphone in a room may be subject to reverberation. Performance of a speaker verification system may degrade significantly for reverberant speech, with severe consequences in a wide range of real applications. This paper presents a comprehensive study of the effect of reverberation on speaker verification, and investigates approaches to reduce the effect of reverberation: training target models with reverberant speech signals and using acoustically matched models for the reverberant speech under test, score normalization methods to improve the reverberation robustness, and also reverberation classification via the background model scores. Experimental investigation is performed, using simulated and measured room impulse responses, NIST-based speech database, and AGMM based speaker verification system, showing significant improvement in performance.
32 citations
01 Mar 2008
TL;DR: An innovative automatic system for detection of elderly people falls at home based on floor vibration and acoustic sensing, and uses pattern recognition algorithm to discriminate between human fall events and other events is presented.
Abstract: Falls are very prevalent among the elderly especially in their home. Approximately one in every three adults 65 years old or older falls each year, 30% of those falls result in serious injuries and more than 70% of the event are at home. Rapid detection of fall events can reduce the rate of mortality and raise the chances to survive the event and return to independent living. In this paper we present an innovative automatic system for detection of elderly people falls at home. The system is based on floor vibration and acoustic sensing, and uses pattern recognition algorithm to discriminate between human fall events and other events. The proposed solution is unique, inexpensive, and does not require the person to wear anything. Using the proposed system we can detect human falls with a sensitivity of 97.5% and specificity of 98.5%.
14 citations
01 Mar 2008
TL;DR: In this paper, the effect of other room parameters such as room dimensions and reflection coefficients of the walls on speaker verification performance was investigated, and the results of SVR with reverberant speech of the same RT were shown to be essentially different.
Abstract: The performance of speaker verification (SVR) systems degrades with the presence of room reverberation. Reverberation results in mismatched conditions between target models and test segments. Reverberation time (RT) is commonly used as a room parameter that represents reverberation. We investigate the effect of other room parameters such as room dimensions and reflection coefficients of the walls on SVR. Equal error rate (EER) is calculated by using room dimensions and reflection coefficients as parameters. Results of SVR with reverberant speech of the same RT are shown to be essentially different, when other room parameters are different. Feature normalization techniques are tested with reverberant speech of the same RT, and shown to be either improving or degrading the performance of SVR when other room parameters are different. This stands in contradiction to the approach in the literature towards RT as a dominant room parameter.
9 citations
01 Mar 2008
TL;DR: Results show that acoustic features from speech signals of awake subjects can predict OSA, and can be used as a tool for initial screening of potential patients.
Abstract: Obstructive sleep apnea (OSA) is a common sleep-related breathing disorder. Previous studies have confirmed that OSA is associated with anatomical abnormalities of the upper airways. Acoustic parameters of human speech are significantly influenced by the vocal tract structure and soft tissue properties; therefore, there is reason to believe that there is correlation between speech signal parameters and the existence of OSA. This work aims to explore the influence of OSA on acoustic speech features. Signal processing and pattern recognition algorithms were developed to differentiate between OSA and non-OSA subjects using their speech signals. Using Gaussian mixture model (GMM) classifier and a speech database of 13 non-OSA and 13 OSA diagnosed adult male subjects, an equal error rate (EER) of 7.7% was achieved. These results show that acoustic features from speech signals of awake subjects can predict OSA, and can be used as a tool for initial screening of potential patients.
9 citations
06 May 2008
TL;DR: The effect of CMS on state of the art GMM and AGMM-based speaker recognition systems is investigated for reverberant speech and results show that high reverberation time reduces the effectiveness of CMS.
Abstract: Speaker recognition is used today in a wide range of applications. The presence of reverberation, in hands-free systems for example, results in performance degradation. The effect of reverberation on the feature vectors and its relation to optimal GMM order are investigated. Optimal model order is calculated in terms of minimum BIC and KIC, and tested for EER of a GMM-based speaker recognition system. Experimental results show that for high reverberation time, reducing model order reduces EER values of speaker recognition. The effect of CMS on state of the art GMM and AGMM-based speaker recognition systems is investigated for reverberant speech. Results show that high reverberation time reduces the effectiveness of CMS.
8 citations
06 May 2008
TL;DR: The definition and centra-time, acoustic parameters which are affected by both room properties and distance, were found to be more correlated with the degradation in the speaker recognition performance.
Abstract: The performance of speaker recognition systems may degrade significantly when speech is recorded in reverberant environments by a microphone positioned far from the speaker. Most of the literature on speaker recognition uses the reverberation time to classify the reverberation effects. However, as described in this work, the reverberation time is mainly a room feature and is less affected by the distance between the source and the microphone. This paper presents a comprehensive study of room acoustics parameters and their relationship with speaker recognition performance. The definition and centra-time, acoustic parameters which are affected by both room properties and distance, were found to be more correlated with the degradation in the speaker recognition performance.
8 citations
01 Mar 2008
TL;DR: In this article, a session variability method at the feature level, which is based on eigenchannel modeling, was used for speaker verification under high reverberation times, and the results showed that this method leads to the best results under reverberation.
Abstract: Reverberation effects introduce a great challenge for speaker verification systems. However, only few studies considered reverberation with speaker verification. These studies did not exploit some of the advanced techniques for channel mismatch in the speaker verification literature. Contemporary techniques for channel and microphones mismatch in speaker verification are based on session variability modeling. One such method is the compensation of nuisance factors, a session variability method at the feature level, which is based on eigenchannel modeling. We adapted this method for reverberant speech and also integrated it with the reverberant model matching method examined in previous studies. The integration leads to the best results under reverberation, especially for high reverberation times.