The role of spatial detail in sound-source localization: Impact on HRTF modeling and personalization.

Romigh, Griffin D.; Brungart, Douglas S.; Stern, Richard M.; Simpson, Brian D.

doi:10.1121/1.4799575

Cited by 4 publications

(6 citation statements)

References 0 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Individualized HRTFs are needed for synthesizing accurate spatial audio that resolve front-back and up-down directional confusion [3], [5], [4]. Due to the difficulties of directly measuring HRTFs [23], a number of works have sought indirect means for learning the subject's HRTFs: regression models between the individual's physically measured anthropometry and his/her HRTFs can be learned via neural-network [24] and multiple non-linear regression models [25] but do not generalize well to test subjects.…”

Section: B Active-learning For Individualizing Hrtfsmentioning

confidence: 99%

“…We propose to formulate the recommendation problem in an active-learning [30] context described as follows: given a finite set of candidate HRTFs X C sampled from a prior distribution (database or generative model), determine the HRTF from the X C that the listener would localize nearest to u within T rounds 5 Evaluates a fitness function w.r.t. localization accuracy of known u of localizations.…”

Section: B Active-learning For Individualizing Hrtfsmentioning

confidence: 99%

“…The spectral ratio between the sounds recorded at the eardrum and that would have been obtained at the center of the head in absence of the listener is known as the head-related transfer function (HRTF) [1]; HRTFs are thus specific to the individual's anthropometry, wave direction, and contain other important cues such as the interaural time delay (ITD) and the interaural level difference (ILD) [2]. Moreover, knowledge of individualized HRTFs allow for perceptually accurate 3D spatial audio synthesis [3], [4], [5].…”

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

Gaussian process models for HRTF based 3D sound localization

Luo

Zotkin

Duraiswami

2014

2014 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP)

View full text Add to dashboard Cite

The human ability to localize sound-source direction using just two receivers is a complex process of direction inference from spectral cues of sound arriving at the ears. While these cues can be described using the well-known head-related transfer function (HRTF) concept, it is unclear as to how densely HRTF must be sampled and whether a higher-order representation is employed in localization. We propose a class of binaural sound source localization models to answer these two questions. First, using the sound received by two ears, we derive several binaural features that are invariant to the sound source signal. Second, these are implicitly mapped to a highdimensional reproducing kernel Hilbert space via a Gaussian process regression model for feature-direction tuples. Lastly, the features that are most relevant in the model are found via an efficient forward subset-selection method. Experimental results are shown for HRTFs belonging to the CIPIC database.

show abstract

Section: B Active-learning For Individualizing Hrtfsmentioning

confidence: 99%

Section: B Active-learning For Individualizing Hrtfsmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Gaussian process models for HRTF based 3D sound localization

Luo

Zotkin

Duraiswami

2014

2014 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP)

View full text Add to dashboard Cite

show abstract

“…Kulkarni and Colburn [22] proposed to truncate the HRTF log-magnitude spectrum after Fourier series expansion, and the resulting smooth HRTF still remains perceptually relevant. On this basis, Romigh et al [23] explored a method for smoothing HRTFs by utilizing a truncated spherical harmonic expansion, and the results showed that the significant smoothing of HRTF in frequency brought by the low-order spherical harmonic representation does not affect the perceived position of the sound.…”

Section: Introductionmentioning

confidence: 99%

Magnitude Modeling of Personalized HRTF Based on Ear Images and Anthropometric Measurements

2022

View full text Add to dashboard Cite

In this paper, we propose a global personalized head-related transfer function (HRTF) method based on anthropometric measurements and ear images. The model consists of two sub-networks. The first is the VGG-Ear Model, which extracts features from the ear images. The second sub-network uses anthropometric measurements, ear features, and frequency information to predict the spherical harmonic (SH) coefficients. Finally, the personalized HRTF is obtained through inverse spherical harmonic transform (SHT) reconstruction. With only one training, the HRTF in all directions can be obtained, which greatly reduces the parameters and training cost of the model. To objectively evaluate the proposed method, we calculate the spectral distance (SD) between the predicted HRTF and the actual HRTF. The results show that the SD provided by this method is 5.31 dB, which is better than the average HRTF of 7.61 dB. In particular, the SD value is only increased by 0.09 dB compared to directly using the pinna measurements.

show abstract

“…It should be noted that this is the most stringent criterion possible to discrimination tasks, which in some other applications might be relaxed [cf. Kulkarni and Colburn (1998) and Romigh et al (2013)]. …”

Section: Introductionmentioning

confidence: 99%

Smoothing individual head-related transfer functions in the frequency and spatial domains

Rasumow

Blau

Hansen

et al. 2014

The Journal of the Acoustical Society of America

View full text Add to dashboard Cite

When re-synthesizing individual head related transfer functions (HRTFs) with a microphone array, smoothing HRTFs spectrally and/or spatially prior to the computation of appropriate microphone filters may improve the synthesis accuracy. In this study, the limits of the associated HRTF modifications, until which no perceptual degradations occur, are explored. First, complex spectral smoothing of HRTFs into constant relative bandwidths was considered. As a prerequisite to complex smoothing, the HRTF phase spectra were substituted by linear phases, either for the whole frequency range or above a certain cut-off frequency only. The results indicate that a broadband phase linearization of HRTFs can be perceived for certain directions/subjects and that the thresholds can be predicted by a simple model. HRTF phase spectra can be linearized above 1 kHz without being detectable. After substituting the original phase by a linear phase above 5 kHz, HRTFs may be smoothed complexly into constant relative bandwidths of 1/5 octave, without introducing noticeable artifacts. Second, spatially smoother HRTF directivity patterns were obtained by levelling out spatial notches. It turned out that spatial notches do not have to be retained if they are less than 29 dB below the maximum level in the directivity pattern.

show abstract

The role of spatial detail in sound-source localization: Impact on HRTF modeling and personalization.

Cited by 4 publications

References 0 publications

Gaussian process models for HRTF based 3D sound localization

Gaussian process models for HRTF based 3D sound localization

Magnitude Modeling of Personalized HRTF Based on Ear Images and Anthropometric Measurements

Smoothing individual head-related transfer functions in the frequency and spatial domains

Contact Info

Product

Resources

About