Filters








14,913 Hits in 6.2 sec

On Open-Set Speaker Identification with I-Vectors

Kevin Wilkinghoff
2020 Odyssey 2020 The Speaker and Language Recognition Workshop   unpublished
In this paper, an open-set speaker identification system based on i-vectors is presented.  ...  The system consists of an outlier detector in combination with a classical closed-set speaker identification chain and utilizes an effective preprocessing technique for i-vectors, called linear alignment  ...  In conclusion, open-set speaker identification is more challenging but also more realistic than closed-set speaker identification.  ... 
doi:10.21437/odyssey.2020-58 fatcat:mgf5djd5ezc4nihqwjuur2m3xm

Toward Open-Set Text-Independent Speaker Identification in Tactical Communications

Matt B. Wolf, WonKyung Park, Jae C. Oh, Misty K. Blowers
2007 2007 IEEE Symposium on Computational Intelligence in Security and Defense Applications  
We present the design and implementation of an open-set textindependent speaker identification system using genetic Learning Classifier Systems (LCS).  ...  We also identify several difficulties in solving the speaker identification problems with LCS and introduce new approaches to resolve the difficulties.  ...  So far, we have trained the system with 20 vectors per speaker for open-set testing. We have conducted experiments using four different sets of data. Table I shows the results.  ... 
doi:10.1109/cisda.2007.368129 dblp:conf/cisda/WolfPOB07 fatcat:ruu6vxc46rfn7orswdroofiyca

Improved Weighted Matching For Speaker Recognition

Ozan Mut, Mehmet Göktürk
2007 Zenodo  
In the case of speaker identification, the result depends on whether the identification is open set or closed set.  ...  In open set identification, the best score is tested against a threshold, so there is one more possible output satisfying the condition that the speaker is not one of the registered speakers in existing  ...  In open-set form on the other hand, the speaker may not belong to one of the registered speakers in the database, therefore an open-set identification system has one more possible output for rejection.  ... 
doi:10.5281/zenodo.1076518 fatcat:2mmjohpu5zblxj6apqlco2xpem

Experiments on Open-Set Speaker Identification with Discriminatively Trained Neural Networks [article]

Stefano Imoscopi, Volodya Grancharov, Sigurdur Sverrisson, Erlendur Karlsson, Harald Pobloth
2019 arXiv   pre-print
This paper presents a study on discriminative artificial neural network classifiers in the context of open-set speaker identification.  ...  Both 2-class and multi-class architectures are tested against the conventional Gaussian mixture model based classifier on enrolled speaker sets of different sizes.  ...  The input to the open-set identification system is a feature vector set X extracted from an audio recording associated with the speaker under test.  ... 
arXiv:1904.01269v1 fatcat:mvihbz3xg5etxmpqzpplbn2yeq

Exploring the Encoding Layer and Loss Function in End-to-End Speaker and Language Recognition System [article]

Weicheng Cai, Jinkun Chen, Ming Li
2018 arXiv   pre-print
In terms of loss function for open-set speaker verification, to get more discriminative speaker embedding, center loss and angular softmax loss is introduced in the end-to-end system.  ...  Experimental results on Voxceleb and NIST LRE 07 datasets show that the performance of end-to-end learning system could be significantly improved by the proposed encoding layer and loss function.  ...  He gives insightful advice on the implementation of end-to-end discriminative loss. This research was funded in part by the National Natural Science  ... 
arXiv:1804.05160v1 fatcat:5ar3oyo23zb5hcnrhozpvpx6cq

Speaker Identification using Spectrograms of Varying Frame Sizes

H. B.Kekre, Vaishali Kulkarni, Prashant Gaikar, Nishant Gupta
2012 International Journal of Computer Applications  
In this paper, a text dependent speaker recognition algorithm based on spectrogram is proposed.  ...  Feature vector extraction has been done by using the row mean vector of the spectrograms. For feature matching, two distance measures, namely Euclidean distance and Manhattan distance have been used.  ...  In the open set identification task, the decision is given as given by eq. (4). reject accept S X dist i i i , , ) , ( (4) Where Θi is the threshold.  ... 
doi:10.5120/7921-1228 fatcat:l5juf2m5ijfdjotwz5tqeqwwse

MCE 2018: The 1st Multi-target Speaker Detection and Identification Challenge Evaluation [article]

Suwon Shon, Najim Dehak, Douglas Reynolds, James Glass
2019 arXiv   pre-print
The task is to measure how accurately one can detect 1) whether a test recording is spoken by a blacklisted speaker, and 2) which specific blacklisted speaker was talking.  ...  It is a form of multi-target speaker detection based on real-world telephone conversations. Data recordings are generated from call center customer-agent conversations.  ...  ://mce.csail.mit.edu/ 4 We also tried a discriminatively trained x-vector embedding using the train set with 5,000 background speakers, but the i-vector system performed better on the MCE task  ... 
arXiv:1904.04240v1 fatcat:t52cze7bizgwna32umgjm4aeki

MCE 2018: The 1st Multi-Target Speaker Detection and Identification Challenge Evaluation

Suwon Shon, Najim Dehak, Douglas Reynolds, James Glass
2019 Interspeech 2019  
The task is to measure how accurately one can detect 1) whether a test recording is spoken by a blacklisted speaker, and 2) which specific blacklisted speaker was talking.  ...  It is a form of multi-target speaker detection based on real-world telephone conversations. Data recordings are generated from call center customer-agent conversations.  ...  Multitarget detection such as blacklist or watchlist was often described as open-set speaker identification.  ... 
doi:10.21437/interspeech.2019-1572 dblp:conf/interspeech/ShonDRG19 fatcat:grgsfm4hffejtnyiinlethry6y

Evaluation of EMD-Based Speaker Recognition Using ISCSLP2006 Chinese Speaker Recognition Evaluation Corpus [chapter]

Shingo Kuroiwa, Satoru Tsuge, Masahiko Kita, Fuji Ren
2006 Lecture Notes in Computer Science  
Since the identification tasks defined in the evaluation were on an open-set basis, we introduce a new speaker verification module in this paper.  ...  The EMD based speaker recognition (EMD-SR) was originally designed to apply to a distributed speaker identification system, in which the feature vectors are compressed by vector quantization at a terminal  ...  Experimental results on the ISCSLP2006 text-independent speaker identification task under the closed-channel condition in the open-set manner, showed that the proposed method achieved 99.33% Identification  ... 
doi:10.1007/11939993_56 fatcat:xcot4fenmvbgnn6s2ccx7korem

A Review on Speaker Recognition

Sujiya S, Dr.Chandra E
2017 International Journal of Engineering and Technology  
On the other hand, identification is the work of determining an unknown speaker's identity.  ...  Therefore to protect one's resources or information confidentially with simple password is not consistent and secure in the technological world of today.  ...  Text Independent Speaker Identification vs. Speaker Verification Open Set vs. Closed Set TABLE I . I Speech Corpus for speaker recognition [18] [19] .  ... 
doi:10.21817/ijet/2017/v9i3/170903513 fatcat:z3pulpp65vaqdmtjlqh75dd52y

OCR-aided person annotation and label propagation for speaker modeling in TV shows

Mateusz Budnik, Laurent Besacier, Ali Khodabakhsh, Cenk Demiroglu
2016 2016 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP)  
At each step of the cycle, annotations are used to build speaker models. The quality of the generated speaker models is evaluated at each step using an i-vector based speaker identification system.  ...  The presented approach shows promising results on the REPERE corpus with a minimum amount of human effort for annotation.  ...  Identification is done by extraction of the i-vector of a target track and calculation of the cosine similarity score of the extracted i-vector with all the speaker models (each represented with an i-vector  ... 
doi:10.1109/icassp.2016.7472743 dblp:conf/icassp/BudnikBKD16 fatcat:k6a7xeihzfflvg4xkfa52svuri

Large-Scale Speaker Retrieval on Random Speaker Variability Subspace

Suwon Shon, Younggun Lee, Taesu Kim
2019 Interspeech 2019  
A recent study shows that Locality Sensitive Hashing (LSH) enables quick retrieval of a relevant voice in the large-scale data in conjunction with i-vector while maintaining accuracy.  ...  We hypothesized that rather than projecting on completely random subspace without considering data, projecting on randomly generated speaker variability space would give more chance to put the same speaker  ...  In the previous study [6] , LSH was adopted for this task and showed very promising performance on the large-scale speaker identification task combining with i-vector.  ... 
doi:10.21437/interspeech.2019-1498 dblp:conf/interspeech/ShonLK19 fatcat:2go6y4cukjhbjjoijdynhi34ky

Large-scale Speaker Retrieval on Random Speaker Variability Subspace [article]

Suwon Shon, Younggun Lee, Taesu Kim
2019 arXiv   pre-print
A recent study shows that Locality Sensitive Hashing (LSH) enables quick retrieval of a relevant voice in the large-scale data in conjunction with i-vector while maintaining accuracy.  ...  We hypothesized that rather than projecting on completely random subspace without considering data, projecting on randomly generated speaker variability space would give more chance to put the same speaker  ...  In the previous study, LSH was adopted for this task and showed very promising performance on the large-scale speaker identification task combining with i-vector.  ... 
arXiv:1811.10812v2 fatcat:qsj45ds63zcphdg6w33xnmrs34

MCE 2018: The 1st Multi-target Speaker Detection and Identification Challenge Evaluation (MCE) Plan, Dataset and Baseline System [article]

Suwon Shon, Najim Dehak, Douglas Reynolds, James Glass
2018 arXiv   pre-print
Each conversation is represented by a single i-vector.  ...  Given a pool of training and development data from non-Blacklist and Blacklist speakers, the task is to measure how accurately one can detect 1) whether a test recording is spoken by a Blacklist speaker  ...  Data format The i-vector and its speaker identification label will be provided in the following CSV format files: trn blacklist.csv trn background.csv dev blacklist.csv dev background.csv tst mix.csv bl  ... 
arXiv:1807.06663v1 fatcat:4sqocdbxxbbqjgadxm7d4essoi

Pindrop Labs' Submission to the First Multi-Target Speaker Detection and Identification Challenge

Elie Khoury, Khaled Lakhdhar, Andrew Vaughan, Ganesh Sivaraman, Parav Nagarsheth
2019 Interspeech 2019  
While one single system can answer both questions, this work looks at them as two separate tasks: blacklist detection and closed-set identification.  ...  Particularly, it aims to answer the following two questions: Is the speaker of the test utterance on the blacklist? If so, which speaker is it among the blacklisted speakers?  ...  However, the task of multi-speaker detection, also known as open-set speaker identification, has received far less attention.  ... 
doi:10.21437/interspeech.2019-3179 dblp:conf/interspeech/KhouryLVSN19 fatcat:ogri7kw43zc7dgq3s7n5jn43ta
« Previous Showing results 1 — 15 out of 14,913 results