Wyniki wyszukiwania - BazTech

1

Logatom articulation index evaluation of speech enhanced by blind source separation and single-channel noise reduction

Drgas Sz., Kociński J., Sęk A. P.

Archives of Acoustics

|

2008

|

Vol. 33, No. 4

455-474

EN

The subjective logatom articulation index of speech signals enhanced by means of various digital signal processing methods has been measured. To improve intelligibility, the convolutive blind source separation (BSS) algorithm by Parra and Spence [1] has been used in combination with classical denoising algorithms. The efficiency of these algorithms has been investigated for speech material recorded in two spatial configurations. It has been shown that the BSS algorithm can highly improve speech recognition. Moreover, a combination of the BSS with single-microphone denoising methods can additionally increase the logatom articulation index.

2

Detection of asynchronicity in the amplitude modulation domain

Kutzner D., Lemańska J., Sęk A. P.

Archives of Acoustics

|

2005

|

Vol. 30, No. 3

291-307

EN

A just noticeable time delay (JNTD) between the onset of a single sinusoidal amplitude modulation (AM) and a complex modulation applied to the same carrier was measured in this study. The carrier was a 4-kHz tone and the modulator was a five-component multitone complex. In the first experiment, four of five components had constant frequencies, i.e. 160, 170, 180, 190 Hz and they were turned on synchronously (synchronous components) in the middle of the carrier duration. The frequency of the fifth component (asynchronous one) varied from 10 to 150 Hz and it was turned on earlier than the synchronous ones. In the second experiment, the asynchronous component was situated in the centre of the synchronous components' spectrum; its frequency was constant and equal to 100 Hz. The spectral separation between the asynchronous component and the synchronous ones of the modulator varied. The results, i.e. the just noticeable time delay between the onset of a single sinusoidal amplitude modulation and a complex modulation (or asynchrony threshold), are analogous to those obtained in the audible frequency domain. They can be interpreted on the basis of the auditory system model containing a bank of modulation filters. It seems that two separate mechanisms are responsible for the JNTD between the onset of the single component modulation and the complex modulation. The first one results from an interaction between all the components of a modulator passing a single modulation filter tuned to the frequency of the asynchronous component. This sort of interaction (or masking) was most effective when the spectral separation between the asynchronous component and the synchronous ones was the smallest one. With an increase in this separation, a significant decrease in the asynchrony thresholds was observed. The second mechanism determining the obtained asynchrony thresholds is based on the uncertainty principle: modulation filters with good frequency selectivity, i.e. filters tuned to low modulation rates, are characterised by a poor time resolution. Thus, in the case of the lowest frequencies of the asynchronous component the subjects' performance would be relatively poor even when there was a significant spectral interval between this component and the synchronous ones. As in the audible frequency domain, the pattern of the asynchronicity thresholds was related to the modulation filter bandwidth. The obtained results suggest the bandwidth of the modulation filters whose Q factor should be close to 1 or less.

3

Masking in the modulation rate domain

Lemańska J., Sęk A. P., Rybicka W.

Archives of Acoustics

|

2003

|

Vol. 28, no. 3

181-189

EN

This study examines the amplitude modulation (AM) detection in the presence of a masking modulating signal. The detection thresholds of sinusoidal amplitude modulation were measured for a 4 kHz sinusoidal carrier. The masking modulators were a 16 Hz tone, a bandwidth of a low-noise noise centered at 16 Hz with a bandwidth of 2 or 8 Hz, and the gaussian noise centered at 16 Hz with a 8 Hz bandwidth. A 3AFC procedure was used. The results obtained suggest the existence of a masking effect in the modulation rate domain. This form of masking is the most effective one when the modulation frequency of the masking signal is close to the masker or, in the spectral range of masking, to the modulation signal. These results are consistent with experimental data, which suggest the existence of frequency selectivity and tuning in the amplitude modulation domain. The results obtained are consistent with the idea of a second stage of filtering in the auditory system by means of so-called modulation filters. It seems that the auditory system performs a limited resolution spectral analysis of the signal amplitude envelope. However, it is necessary to stress that the frequency selectivity in the amplitude modulation domain is not so evident as the selectivity in the audio frequency domain.

4

Discrimination of the amplitude modulation rate

Lemańska J., Sęk A. P., Skrodzka E. B.

Archives of Acoustics

|

2002

|

Vol. 27, no. 1

3-22

EN

This study examines the amplitude modulation rate discrimination for sinusoidal and noise carriers. It was shown that the discrimination of AM rates is a monotonically growing function of modulation rate. Higher values of the discrimination thresholds were observed for a narrowband carrier. It appears that in the case of a narrowband noise carrier, the spectral range of the noise envelope is similar to that of the modulation rates of the signal (up to 120 Hz). It results in a masking in the modulation rate domain and in a much higher threshold growth than that observed for a wideband noise carrier or a sinusoidal carrier. The results are consistent with the idea of the so-called second stage of filtering acting on the envelope of the acoustic signal. This hypothesis postulates the existence of a so-called modulation filter bank, (MFB), responsible for the frequency selectivity observed in the amplitude modulation rate domain. The existence of the MFB suggests that a certain form of the spectral analysis of any acoustic signal envelope may be performed in the auditory system after initial filtering in the auditory filter bank. A model of the modulation rate discrimination based either on the classical concept of the excitation patterns or on the modulation excitation patterns has not accounted for our experimental data. According to both the models, an increase in the frequency discrimination threshold versus modulation rate should be slower than that measured in the experiment.

5

Auditory filters in sensorineural hearing impaired subjects

Skrodzka E. B., Wicher A., Ozimek E., Sęk A. P.

Archives of Acoustics

|

2002

|

Vol. 27, no. 3

159-174

EN

The study is devoted to determination of the shape of the auditory filters in subjects with sensorineural hearing loss. Apart from the classical sensorineural hearing loss, changes in the auditory filter shapes have been analysed in the subject diagnosed with dead regions. The dead region is an area on the basilar membrane over which the functioning of the inner hair cells and/or neurones innervating them has ceased. This type of hearing impairment means that the information on the sounds whose frequencies correspond to the dead region of the basilar membrane are to a very limited degree if at all, transmitted to higher levels of the auditory path. This transmission, if happens, is performed through the auditory filters at the centre frequency other than that of the signal. This phenomenon and the fact that in the dead region the hearing loss is theoretically infinite, affect the transmittance of the auditory filters. Results of the study reported here have shown that in general, the subjects with sensorineural hearing loss develop broadening of the auditory filter accompanied by reducing of its dynamics. This fact explains a considerable decrease in speech intelligibility presented at a background of a noise. In the subject with the dead regions the broadening of the filters was the greatest in the region of the dead one. The results also indicate the lack of correlation between the shape and width of the auditory filter and the shape of the audiogram.