Flemming Christensen

Quantitative assessment of spatial sound distortion by the semi-ideal recording point of a hear-through device

A hear-through device combines a microphone and earphone in an earpiece so that when worn, one per ear, it can work as an acoustically transparent system allowing for simultaneous individual binaural recording and playback of the real sound field at the ears. Recognizing the blocked entrance to the ear canal as the ideal recording point - i.e. all directional properties of the incident sound field are recorded without distortion - it is critical for such device to be sufficiently small so that it can be completely inserted into the ear canal. This is not always feasible and the device may stretch out from the ideal position and thus distort the captured spatial information. Here we present measurements that quantify by how much the directional properties of the sound field are distorted by semi-ideal hear-through prototypes built by mounting miniature microphones on the outer part of selected commercial insert earphones. This includes an analysis of the magnitude by which spatial information is distorted ...

Localization with artificial head recordings

Twelve artificial heads, of which one was designed at the authors’ laboratory, were evaluated in a localization test. The localization performance was compared in two situations: First, the subjects localized sound sources in a real sound field, then the localization test was repeated with artificial head recordings of the same sound field. The sounds to be localized were loudspeaker reproductions of female speech at natural level, from 19 different positions in a standard listening room. The artificial head recordings were reproduced by carefully equalized headphones. Results from eight of the heads showed an increased number of localization errors compared to real‐life performance. The directions in the median plane were most often confused, not only with nearby directions, but also with directions further away. The number of errors was significantly higher than what can be obtained with recordings from a carefully selected human head. Results from the last four heads, including the one from the authors’ laboratory, were not available at the time of abstract submission.

Identifying the dominating perceptual differences in headphone reproduction

The perceptual differences between the sound reproductions of headphones were investigated in a pair-wise comparison study. Two musical excerpts were reproduced over 21 headphones positioned on a mannequin and recorded. The recordings were then processed and reproduced over one set of headphones to listeners, who were asked to evaluate their perceived degree of dissimilarity. The two musical excerpts were used in separate experiments. The processing of the recordings consisted of compensating for the influences of the playback headphones worn by the listeners as well as for the mannequins ear canals. A multidimensional scaling analysis revealed two dominating perceptual dimensions used by the listeners to differentiate the reproductions of the headphones. These dimensions were similar for the two musical excerpts. Objective metrics are proposed to describe them, leading to correlations ranging from 0.89 to 0.97 between the dimensions and metrics. The first perceptual dimension was associated with the relative strength of bass, while the second dimension was related to the relative strength of the lower midrange.

Determination of optimal HRTFs for binaural synthesis

Earlier studies have shown that a carefully selected human substituting an artificial head can minimize the amount of localization errors for a group of people in a listening experiment, thus indicating that a set of head‐related transfer functions (HRTFs) can be found, which to some extent fits a population. This study aims at exploring ways of designing such general sets of HRTFs suitable for larger populations. An effort is put into considering the importance of, e.g., different frequency regions of the transfer functions in order to focus on the most general characteristics and avoid focusing attention on highly individual features. The physical origin of the different parts of the HRTFs will be taken into consideration, and HRTF design methods using parameters derived through signal analysis will be studied.

Human localization of sound signals with reduced bandwidth

In multimedia systems, the sound is often presented with reduced bandwith. As 3‐D sound is introduced, this might give rise to localization problems, a consequence which is partly confirmed by this investigation. A listening experiment was made, in which 12 subjects participated. The subjects were placed in an anechoic room in a setup with 17 loudspeakers placed in different directions. As a reference experiment the subjects listened to a pink noise signal played from one loudspeaker at a time. The subjects were asked to point out the loudspeaker from which they perceived the sound. In eight other experiments the sound was low‐pass filtered at frequencies of 1, 2, 4, or 8 kHz, or high‐pass filtered at 0.5, 1, 2, or 4 kHz. The procedure was the same as in the reference experiment. For each subject the order of the tests was randomized. The experiment showed that all four low‐pass filtered signals gave an increase in localization errors. This was not the case with the high‐pass filtered signals.