Human Intention Inference Using Expectation-Maximization Algorithm With Online Model Learning

Ravichandar, Harish; Dani, Ashwin P.

doi:10.1109/tase.2016.2624279

Cited by 47 publications

(31 citation statements)

References 41 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In order to obtain the best GMM parameter λ, that is, the distribution of the concerned voice features has the greatest similarity to the distribution of the model parameter λ, it is necessary to estimate the most suitable model parameter λ, the probability density of which can be expressed as follows 38 : The vector distribution of these 128 groups is the initial parameter (ie, feature vector) distribution of the GMM.…”

Section: Em Algorithmmentioning

confidence: 99%

“…In order to find the model parameter λ′ that can maximize the likelihood function value of the GMM, as mentioned above, we use the EM algorithm 38 to interactively find the Gaussian models of the GMM. In order to find the model parameter λ′ that can maximize the likelihood function value of the GMM, as mentioned above, we use the EM algorithm 38 to interactively find the Gaussian models of the GMM.…”

Section: Em Algorithmmentioning

confidence: 99%

“…The vector distribution of these 128 groups is the initial parameter (ie, feature vector) distribution of the GMM. In order to obtain the best GMM parameter λ, that is, the distribution of the concerned voice features has the greatest similarity to the distribution of the model parameter λ, it is necessary to estimate the most suitable model parameter λ, the probability density of which can be expressed as follows 38 :…”

Section: Em Algorithmmentioning

confidence: 99%

“…in which X i is one of the 128 Gaussian models after clustering and X ={X 1 , X 2 , … , X K } are the feature vectors of GMM that is deterministic. In order to find the model parameter λ′ that can maximize the likelihood function value of the GMM, as mentioned above, we use the EM algorithm 38 to interactively find the Gaussian models of the GMM. In the first iteration, the EM algorithm re-estimates the new model parameter λ′ by using the initialization parameter λ obtained by invoking the K-means clustering method so that P(X| λ′) ≥ P(X| λ).…”

Section: Em Algorithmmentioning

confidence: 99%

“…The similarities and identification accuracies of Experiment 1 performed on the words listed inTable 2as the training and test data (%)38.15 47.05 45.81 24.96 70.44 55.42 79.43 64.76 44.28 36.97 28.21 41.52 25.01 34.40 90.65 2 38.70 93.76 69.65 11.23 33.82 28.27 52.06 48.63 44.54 58.29 34.41 48.43 70.48 27.72 39.96 87.37 3 44.20 64.68 94.36 78.19 44.27 7.11 45.86 58.18 65.25 48.94 48.48 40.40 50.91 23.67 18.86 Forty test words used in Experiment 2 (they are different from those listed inTable 2)…”

mentioning

confidence: 99%

See 4 more Smart Citations

An MFCC‐based text‐independent speaker identification system for access control

Liu

Leu

Lin

et al. 2017

Concurrency and Computation

View full text Add to dashboard Cite

In recent years, by merit of convenient and unique features, bio-authentication techniques have been applied to identify and authenticate a person based on his/her spoken words and/or sentences. Among these techniques, speaker recognition/identification is the most convenient one, providing a secure and strong authentication solution viable for a wide range of applications.In this paper, to safeguard real-world objects, like buildings, we develop a speaker identification system named mel frequency cepstral coefficients (MFCC)-based speaker identification system for access control (MSIAC for short), which identifies a speaker U by first collecting U's voice signals and converting the signals to frequency domain. An MFCC-based human auditory filtering model is utilized to adjust the energy levels of different frequencies as U's voice quantified features. Next, a Gaussian mixture model is employed to represent the distribution of the logarithmic features as U's specific acoustic model. When a person, eg, x, would like to access a real-world object protected by the MSIAC, x's acoustic model is compared with known-people's acoustic models. Based on the identification result, the MSIAC will determine whether the access will be accepted or denied. KEYWORDS acoustic model, Fourier transformation, Gaussian mixture model, mel frequency cepstral coefficients, speaker identification 1 | INTRODUCTIONIn this information era, numerous high-tech products gradually enter our everyday lives and significantly change our living habits and patterns. The biometrics identification technology that provides us with easier and more convenient methods to identify people has gradually replaced some existing authentication techniques, which use passwords or pin numbers to authenticate users. But the passwords or pin numbers may be forgotten or forged and are no longer considered to offer a high level of security. The face recognition systems used at airport halls 1 and the voice assistant SIRI of iPhone 2 are two examples of the biometric identification systems.On the one hand, voice has been the most direct and natural method for us to express ideas, communicate with others, and do something for interaction. Therefore, recognizing people's identities from user's dialogue voice and contents and then providing the corresponding services should be a better method to practically make our daily lives easier. Up to present, speech recognition technology 3 has been well developed and applied to our living activities. But speaker recognition technology 4 is still far away from its practical applications. The reasons are that (i) there are too many parameters needed to be processed for speaker recognition; (ii) it is hard to collect voice features completely; and (iii) the identification process is complicated and takes a long time for calculation. Thus, it is difficult to be applied to those applications that need immediate response. Furthermore, the studies of speaker identification nowadays are partial, rather than whole. For example, the Hidden ...

show abstract

Section: Em Algorithmmentioning

confidence: 99%

Section: Em Algorithmmentioning

confidence: 99%

Section: Em Algorithmmentioning

confidence: 99%

Section: Em Algorithmmentioning

confidence: 99%

mentioning

confidence: 99%

See 3 more Smart Citations