Analysis and Recognition of NAM Speech Using HMM Distances and Visual Information

Heracleous, Panikos; Tran, Viet-Anh; Nagai, Takayuki; Shikano, Kiyohiro

doi:10.1109/tasl.2009.2037398

Cited by 13 publications

(12 citation statements)

References 27 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…Said work shows that the system's performance improves, particularly under noise conditions. In addition, other authors [13] use information extracted from acoustic waves travelling through the body tissue of people when speaking, whose signals are picked up by special microphones placed behind the ear. Likewise, several methods have been developed seeking to include visual information about lip movement to improve recognition systems.…”

Section: Introductionmentioning

confidence: 99%

Phoneme recognition system using articulatory-type information

2015

View full text Add to dashboard Cite

This work is frameworked within the development of phoneme recognition systems and seeks to establish whether the incorporation of information related to the movement of the articulators helps to improve the performance thereof. For this purpose, a pair of systems is compared and developed, where the acoustic model is obtained from training hidden Markov chains. The first system represents the voice signal by Mel Frequency Cepstral Coefficients; the second uses the same Cepstral coefficients but together with articulatory parameters. The experiments were conducted on the MOCHA-TIMIT database. The results show a significant increase in the system´s performance by adding articulatory parameters compared to that based only on Mel Frequency Cepstral Coefficients.Keywords: Mel-Cepstrum coefficients, hidden Markov models, articulatory parameters, phoneme recognition ResumenEl presente trabajo se enmarca dentro del desarrollo de sistemas de reconocimiento de fonemas y busca establecer si la incorporación de información relacionada con el movimiento de los articuladores ayuda a mejorar el desempeño de los mismos. Para ello, se desarrollan y comparan un par de sistemas, donde el modelo acústico se obtiene a partir del entrenamiento de cadenas ocultas de Markov. El primer sistema representa la señal de voz mediante coeficientes cepstrales en la escala Mel; y el segundo, utiliza los mismos coeficientes cepstrales pero en conjunto con parámetros articulatorios. Los experimentos fueron realizados sobre la base de datos MOCHA-TIMIT. Los resultados muestran un incremento significativo en el desempeño del sistema al agregar parámetros articulatorios con respecto a sistemas basados en coeficientes cepstrales en la escala Mel.Palabras clave: coeficientes cepstrales en la escala de Mel, modelos ocultos de Markov, parámetros articulatorios, reconocimiento de fonemas

show abstract

Section: Introductionmentioning

confidence: 99%

Phoneme recognition system using articulatory-type information

2015

View full text Add to dashboard Cite

show abstract

“…Bu çalışmada koşula baglı olasılık yogunluk fonksiyonu, Gauss karışım modeli kullanılarak oluşturulmuştur. Ayrıca egitim verisini azaltma çalışmaları Heracleous ve ekibi tarafından yapılmış olup, mırmır mikrofonlarının daha az izgesel alan kaplamasından faydalanarak saklı Markov modelinin (SMM) de normal konuşmaya göre daha az izgesel alan kapladıgını göstermişlerdir [3].…”

Section: Introductionunclassified

A phonetic classification for throat microphone enhancement

Turan

Erzin

2014

2014 22nd Signal Processing and Communications Applications Conference (SIU)

View full text Add to dashboard Cite

Özetçe -Bu analiz çalışmasında gırtlak mikrofonundan kaydedilmiş seslerin izgesel bürüm eşleştirilmesi aracılıgıyla, söyleniş yeri ve biçimine dayalı fonetik sınıflandırılmasındaki etkisini inceledik. Bu sınıflandırmada Gauss karışım modeli (GKM) tabanlı izgesel bürüm eşleştirilmeşemalarını, dogrusal kestirim modelindeki yansıma katsayısı gösterimi kullanarak inceledik. Yansıma katsayılarının, ses üretim yolunu modelleyen kayıpsız tüplerin baglanmaşemasındaki eşleştirme performansı ile ilintili oldugu bilinmektedir. Deneysel çalışmalarda eşleştirilen yansıma katsayılarının logaritmik izgesel bozulumunu ve ortalama karesel hatalarını kullanarak, söyleniş yeri ve biçiminin kümelemeler dahilindeki izgesel eşleştirme başarımını degerlendirdik. Ortaya çıkan sonuçlar, izgesel eşleştirmedeki en büyük bozulmaların söyleniş yerlerine göre süreksiz ve akıcı ünsüzlerde, söyleniş biçimine göre ise yuvarlak ve dişeti ünsüzlerinde oldugunu göstermiştir. Ayrıca söyleniş biçimlerine göre yapılan kümelemeler söyleniş yerlerine göre yapılanlara kıyasla daha yüksek iyileştirme başarımına sahiptir. Anahtar Kelimeler-gırtlak mikrofonu, ses iyileştirmesiAbstract-In this analysis paper, we investigate the effect of phonetic clustering based on place and manner of articulation for the enhancement of throat-microphone speech through spectral envelope mapping. Place of articulation (PoA) and manner of articulation (MoA) dependent GMM-based spectral envelope mapping schemes have been investigated using the reflection coefficient representation of the linear prediction model. Reflection coefficients are expected to localize mapping performance within the concatenation of lossless tubes model of the vocal tract. In experimental studies, we evaluate spectral mapping performance within clusters of the PoA and MoA using the log-spectral distortion (LSD) and as function of reflection coefficient mapping using the mean-square error distance. Our findings indicate that highest degradations after the spectral mapping occur with stops and liquids of the MoA, and velar and alveolar classes of the PoA. The MoA classification attains higher improvements than the PoA classification.

show abstract

“…Segmenting speech signals to its phonemes help recognizer to identify a complete set of words in any language despite the fact that each language just contains tens of phonemes. Older segmentation systems used Dynamic Time Warping (DTW) as classifier [2] [3] [14] while recent methods use Hidden Markov Models (HMMs); consequently, they have achieved more efficiency [4] [5] [15].…”

Section: Introductionmentioning

confidence: 99%

Automatic Speech Segmentation Based On Audio and Optical Flow Visual Classification

Torabi

Nilchi

2014

IJIGSP

View full text Add to dashboard Cite

Automatic speech segmentation as an important part of speech recognition system (ASR) is highly noise dependent. Noise is made by changes in the communication channel, background, level of speaking etc. In recent years, many researchers have proposed noise cancelation techniques and have added visual features from speaker's face to reduce the effect of noise on ASR systems. Removing noise from audio signals depends on the type of the noise; so it cannot be used as a general solution. Adding visual features improve this lack of efficiency, but advanced methods of this type need manual extraction of visual features. In this paper we propose a completely automatic system which uses optical flow vectors from speaker's image sequence to obtain visual features. Then, Hidden Markov Models are trained to segment audio signals from image sequences and audio features based on extracted optical flow. The developed segmentation system based on such method acts totally automatic and become more robust to noise.

show abstract

Analysis and Recognition of NAM Speech Using HMM Distances and Visual Information

Cited by 13 publications

References 27 publications

Phoneme recognition system using articulatory-type information

Phoneme recognition system using articulatory-type information

A phonetic classification for throat microphone enhancement

Automatic Speech Segmentation Based On Audio and Optical Flow Visual Classification

Contact Info

Product

Resources

About