International Journal of Functional Informatics and Personalised Medicine, 2009
... a lot of difficulty to normally process and recognise the speaker-independent isolated ... CK... more ... a lot of difficulty to normally process and recognise the speaker-independent isolated ... CK (1991) Implementation of Hidden Markov Models for Speech Recognition, Thesis, Southern ... Ververidis, D., Kotropoulos, C. and Pitas, I. (2004) 'Automatic emotional speech classification ...
In this work, the circular hidden Markov models (CHMMs) have been used to enhance the recognition... more In this work, the circular hidden Markov models (CHMMs) have been used to enhance the recognition performance of isolated-word text-dependent speaker authentication systems under the neutral talking condition. Our results show that CHMMs enhance the speaker authentication performance under such a condition compared to the left-to-right hidden Markov models (LTRHMMs). The average speaker authentication performance based on using CHMMs has been noticeably improved compared to that based on using LTRHMMs.
Speakers usually use certain words more frequently in expressing their emotions since they have l... more Speakers usually use certain words more frequently in expressing their emotions since they have learned the connection between certain words and their corresponding emotions. This work focuses on speaker-dependent and text-dependent emotion identification in completely two separate and different speech databases. One database uses neutral sentences that are unbiased towards any emotion; however, the second database uses certain sentences that
In this work we focus on Emarati speaker identification systems in neutral talking environments b... more In this work we focus on Emarati speaker identification systems in neutral talking environments based on each of Vector Quantization (VQ), Gaussian Mixture Models (GMMs), and Hidden Markov Models (HMMs) as classifiers. These systems have been tested on our collected Emarati speech database which is composed of 25 male and 25 female Emarati speakers using Mel-Frequency Cepstral Coefficients (MFCCs). Our results yield an average text-dependent Emarati speaker identification performance of 100.00%, %99.81, and 99.69% based on VQ, GMMs, and HMMs, respectively. For text-independent systems, the average Emarati speaker identification performance based on VQ, GMMs, and HMMs is 94.48%, 86.55%, and 74.83%, respectively. The achieved results based on VQ are close to those obtained in subjective assessment by human listeners.
This work is devoted to proposing, implementing and evaluating a two-stage approach to verify spe... more This work is devoted to proposing, implementing and evaluating a two-stage approach to verify speakers in emotional environments using their emotions (emotion-dependent speaker verification problem) based on Hidden Markov Models (HMMs). The results of this work show that verifying speakers from their emotions gives promising results with a significant improvement over emotion-independent speaker verification. The emotional environments in this work
International Journal of Functional Informatics and Personalised Medicine, 2009
... a lot of difficulty to normally process and recognise the speaker-independent isolated ... CK... more ... a lot of difficulty to normally process and recognise the speaker-independent isolated ... CK (1991) Implementation of Hidden Markov Models for Speech Recognition, Thesis, Southern ... Ververidis, D., Kotropoulos, C. and Pitas, I. (2004) 'Automatic emotional speech classification ...
In this work, the circular hidden Markov models (CHMMs) have been used to enhance the recognition... more In this work, the circular hidden Markov models (CHMMs) have been used to enhance the recognition performance of isolated-word text-dependent speaker authentication systems under the neutral talking condition. Our results show that CHMMs enhance the speaker authentication performance under such a condition compared to the left-to-right hidden Markov models (LTRHMMs). The average speaker authentication performance based on using CHMMs has been noticeably improved compared to that based on using LTRHMMs.
Speakers usually use certain words more frequently in expressing their emotions since they have l... more Speakers usually use certain words more frequently in expressing their emotions since they have learned the connection between certain words and their corresponding emotions. This work focuses on speaker-dependent and text-dependent emotion identification in completely two separate and different speech databases. One database uses neutral sentences that are unbiased towards any emotion; however, the second database uses certain sentences that
In this work we focus on Emarati speaker identification systems in neutral talking environments b... more In this work we focus on Emarati speaker identification systems in neutral talking environments based on each of Vector Quantization (VQ), Gaussian Mixture Models (GMMs), and Hidden Markov Models (HMMs) as classifiers. These systems have been tested on our collected Emarati speech database which is composed of 25 male and 25 female Emarati speakers using Mel-Frequency Cepstral Coefficients (MFCCs). Our results yield an average text-dependent Emarati speaker identification performance of 100.00%, %99.81, and 99.69% based on VQ, GMMs, and HMMs, respectively. For text-independent systems, the average Emarati speaker identification performance based on VQ, GMMs, and HMMs is 94.48%, 86.55%, and 74.83%, respectively. The achieved results based on VQ are close to those obtained in subjective assessment by human listeners.
This work is devoted to proposing, implementing and evaluating a two-stage approach to verify spe... more This work is devoted to proposing, implementing and evaluating a two-stage approach to verify speakers in emotional environments using their emotions (emotion-dependent speaker verification problem) based on Hidden Markov Models (HMMs). The results of this work show that verifying speakers from their emotions gives promising results with a significant improvement over emotion-independent speaker verification. The emotional environments in this work
Uploads