PERCEPTION OF SPECTRAL RIPPLES & SPEECH IN NOISE BY GERIATRIC LISTENERS
AUTHORS
Vikas M.D
II MASLP
Department of Audiology and Speech Language Pathology
KMC (A Unit of MANIPAL University), Mangalore-1
ISHA membership no: L -1486
E-mail:
Hassan P. C.
II MASLP
Department of Audiology and Speech Language Pathology
KMC (A Unit of MANIPAL University), Mangalore-1
ISHA membership no: L-1564
e-mail:
Mr. Arivudai Nambi
Assistant Professor in Audiology
Department of Audiology and Speech Language Pathology
KMC (A Unit of MANIPAL University), Mangalore-1
ISHA membership no: L-1009
e-mail:
Paper Submitted for the 43nd ISHACON
PERCEPTION OF SPECTRAL RIPPLES & SPEECH IN NOISE BY GERIATRIC LISTENERS
ABSTRACT
INTRODUCTION
Difficulty in understanding speech especially in noise is one of the most frequent complaints of geriatrics in spite of having normal hearing. This difficulty in speech perception may be related to poor frequency or temporal resolution. Several investigators have studied the frequency resolution by estimated auditory filter shape using notch noise method in geriatrics. They have found that geriatric individuals have broader auditory filters (Patterson et al., 1928; van Rooji et al. 1982). In contrast, Sommers and Humes (1993) reported that auditory filter shape in young and elderly listeners with normal hearing are identical. Recent studies have showed that broadening of auditory filter is associated with hearing impairment but for normal hearing individuals the auditory filter shape did not show any differences (Peters and Moore, 1992).
NEED FOR THE STUDY
There is no clear cut evidence available in literature regarding the spectral resolution and the speech identification abilities in geriatrics with normal hearing. Hence there is a need to study the frequency resolution abilities in geriatrics. There are several methods to measure the frequency resolution like notch noise method, psychophysical tuning curves. All these procedures have discrete steps, and have to cover larger frequency range and hence time consuming. However there are simpler and less time consuming methods like the spectral ripple method to measure the frequency resolution. In this method complex signal is created with ripples (peaks and dips) in spectral domain. Ripple density is increased by increasing the no of ripples per octave. Individual is presented with two stimuli, one with standard spectrum and other with reverse spectrum. Reverse spectrum will have peaks where the standard spectrum had dips and vice versa. Ripple discrimination threshold (RDT) is estimated by measuring highest ripple density at which the individual can differentiate the ripples with standard & reverse spectrum. Current study assesses the frequency resolution by estimating the ripple discrimination threshold.
AIM
1. To measure the RDT in geriatrics and young adults.
2. To investigate the speech recognition scores of geriatrics and young adults in quiet and in the presence of background noise.
3. To correlate the spectral ripple resolution and speech recognition scores in quiet and in noise.
METHOD
Participants for the study included 15 geriatrics of age range between 45 to 61yrs (mean age of 53.04yrs) and 15 young adults with the age range of 18yrs to 25yrs (mean age of 22yrs) with hearing sensitivity within normal limits (within 25dBHL pure-tone thresholds from 250 to 8000 Hz) and no history of any middle ear disorders and neurological disorders.
Signal processing
Stimuli with spectral ripples were created in MATLAB 7 environment. Two-hundred pure-tone frequency components were summed to generate the rippled noise stimuli. The amplitudes of the components were determined by a full-wave rectified sinusoidal envelope on a logarithmic amplitude scale. The ripple peaks were spaced equally on a logarithmic frequency scale. The ripple stimuli were generated with 6 different densities, measured in ripples per octave. The ripple densities differed by ratios of 1.414 (2.000, 2.828, 4.000, 5.657, 8.000, and 11.314 ripples/octave). For standard ripples, the phase of the full-wave rectified sinusoidal spectral envelope was created using ‘sin’ function and for inverted ripples, it was ‘cos’ function.
Four lists of sentences were taken from standardized quick SIN test in Kannada (Kumar, 2009) in order to assess the speech recognition ability of the participants. Each list contained 7 sentences and each sentence has 5 key words, so total of 35 keywords in each list. Speech materials were spoken by male speaker with Indian English accent. The stimuli were recorded digitally on a data acquisition system. Four talker babble was added to 3 lists at +10dB, 0dB SNR & -10dB SNR. List four was presented in quiet.
Procedure
The experiment was performed on a PC equipped with a Creative Labs SoundBlaster 16 soundcard. The subjects listened to the stimuli via Senheiser stereo headphones at a comfortable level set by the subjects.
Classical method of constant stimuli with 2AFC task was used to measure spectral ripple thresholds.10 trails were presented at each ripple density. Subjects heard two intervals, target interval contained standard & reverse ripples and the other interval contained either standard & standard or reverse & reverse ripples. Subject has to identify the target interval by clicking on appropriate button appearing on screen.
Speech recognition was assessed in open set paradigm for all the four condition (quiet, +10dBSNR, 0dBSNR & -10dBSNR). Written responses were taken from each subjects The responses were scored using the ‘loose method’ (Rosen & Faulkner,1992) in which sentence recognition scores were calculated based on correctly identified key words.
RESULTS:
To estimate the ripple discrimination threshold (RDT), probability of correct responses was calculated which followed the weibull distribution. Thresholds were determined using 75% criteria on weibull psychometric function. Independent t-test was administered to investigate the effect of age on RDT. Statistical analysis revealed that young adults performed significantly better (t=-12.031, p<0.05) when compared to geriatrics.
Speech recognition scores in each condition were calculated by counting the number of correctly identified key words. Total scores for each condition were converted into Rationalized Arcsine Unit (RAU) scores to account for the critical differences that inherently present in conventional scoring method (Studebaker, 1985). These RAU scores were used for further statistical analysis. Independent ‘t’ test revealed that speech identifications were significantly different between young adults & geriatrics at 0dB (t=-21.795, p<0.05) & -10dB (t=-3.115, p<0.05) scores. Ceiling was reached for both groups at quiet & +10dB condition. Pearson’s correlation analysis revealed significant positive correlation between ripple discrimination threshold and speech recognition scores at 0dBSNR(r=0.88, p<0.05) and -10dBSNR(r=0.52, p<0.05).
DISCUSSION:
Results of RDT reveals that frequency resolving ability of geriatric individuals is poor when compared to young adults even though they have normal hearing. This difference can be attributed the age related changes in cochlear processing (Patterson et al., 1982) and phase locking (Clinard et.al. 2010). There was no statistical significant difference observed between two groups for +10dB SNR and no noise condition as the responses of both the groups reached the ceiling effect. However, geriatric individual’s had significantly lower speech identification scores at less favourable SNRs i.e., 0 dB SNR and -10 dB SNR. This difficulty in understanding speech in noisy situations can be attributed to the poor frequency resolution as estimated by ripple discrimination threshold. Segregation of target speech and background noise to separate perceptual streams largely depends on frequency resolution ability. Good correlation obtained between RDT and speech in noise also supports the fact that frequency resolution is important to perceive speech in noise.
CONCLUSION
The current study emphasizes the need to study the frequency resolution ability of a geriatric individual in order to study the problems faced by the geriatrics especially during difficult to hear situations. It can be concluded from the current study that poor speech perception by geriatric individual may also be due to poor frequency resolution.