Project Topics
Get Complete Project Material File(s) Now! »
Global loudness of long time-varying sounds
Some studies investigated global loudness of time-varying sounds using long and complex sequences lasting from a few seconds to a few minutes (cf. Fig. 3 (c)) often made of uncontrolled stimuli such as recordings of natural sounds (e.g., Kuwano and Namba 1985; Hellbrück 2000; Gottschling 1999).
There are only a few exceptions where long synthesized sequences and thus “controlled” long time-varying sound sequences were employed (e.g., Susini et al. 2007) (see also Chapter 3). All these studies provided converging evidence that memory processes such as “recency effects” (e.g., Susini et al. 2002, Susini et al. 2007) seem to be involved in global loudness judgments of long time-varying sounds, but we are still far from a full comprehension of the parameters that govern these processes. The advantage of using such long and complex sequences is that the findings of these experiments can be directly interpreted in the context of everyday sound perception and evaluation.
However, the main disadvantage is that such complex stimuli might recruit various mechanisms, so they are probably not the best candidates for psychophysical experiments that aim to disentangle, identify and scrutinize the underlying mechanisms.
Sounds with rising vs. falling amplitude envelopes: Perceptual “asymmetries” at different timescales
The last two or three decades have seen a growing interest in the psychophysical community for what we will call in this thesis “perceptual asymmetries”.
This term refers to the fact that auditory stimuli with the same overall physical characteristics (i.e., level and spectral characteristics) but different asymmetric amplitude envelopes are often perceived differently.
This phenomenon has been considered in the case of short-duration stimuli (ranging from 20 ms to 250 ms) with asymmetrical amplitude envelopes. In particular, ramped or damped stimuli made of exponential increase or decrease in amplitude (cf. Fig. 3 (a)) were found to be with different loudness20 (e.g., Stecker and Hafter 2000), timbre (e.g., Irino and Patterson 1996) and duration (e.g., Schlauch et al. 2001) (see Chapter 1 for a more detailed review). Different mechanisms were proposed to account for these effects, but some pieces of the puzzle are still missing to understand the whole process (see Ries et al. 2008). It is interesting to note that these short asymmetric stimuli induce perceptual “asymmetries” not only concerning loudness, but also other auditory dimensions. This suggests that an intrinsic component of the auditory system might processes sounds having asymmetric amplitude envelopes differently, thus causing consistent perceptual asymmetries between sounds whose amplitude is either rising or falling. Recent studies indicate that these effects would not only involve low-level peripheral mechanisms but also more central stages of auditory processing (e.g., Lentz and Shen 2011). It should be noted however, that although damped stimuli can be compared to impulsive sounds produced in reverberant environments or to musical notes of struck string instruments such as those produced by a piano (see Schutz and Vaisberg 2014), ramped stimuli cannot be found in daily life; they merely correspond to damped sounds played backward. Therefore, we argue that, even such stimuli are particularly good candidates to recruit and thus to investigate nonlinear auditory mechanisms, the outcome of these psychophysical investigations cannot be transposed directly to what occurs with natural sounds.
Investigating intensity coding as a complementary approach
Investigating sound intensity coding at the different stages of the auditory system might also be considered a complementary approach to understanding the mechanisms that underlie loudness perception. Indeed, studying intensity coding can help to clear up the relationships inside the “black box” (inferred from psychophysical experiments) and relate them with biomechanical, physiological and neural properties of the auditory system. Such studies were mostly conducted with stationary sounds and have characterized the main and fundamental functions of sound level coding in the auditory system, from low-level peripheral stages to high-level cortical stages (see Clarey et al. 1992, pp. 245-251; Moore 2003; Schreiner and Malone 2015). However, the relationship between loudness, i.e. a “psychological” quantity, and physiological measurements that can be made along the auditory pathway is still a sensitive issue (Langers et al. 2007; Schreiner and Malone 2015). Only a small number of studies examined neural coding and processing with timevarying sounds (e.g., Lu et al. 2001; Seifritz et al. 2002), and almost nothing is known about the neural basis of global loudness perception of sounds that vary over a few seconds. A better understanding of the cortical processing of time-varying stimuli would certainly constitute a worthwhile asset to determine the temporal mechanisms that can be highlighted in psychophysical studies.
Predictions from current loudness models
It is interesting to compare the magnitude of this asymmetry with predictions from two loudness models applicable to time-varying stimuli (DLM, Chalupper and Fastl 2002; TVL, Glasberg and Moore 2002). Peak values of both short-term (proposed in DLM and TVL) and long-term (in TVL only) loudness time-series predicted by these models were used to evaluate the global loudness of our stimuli. Then, loudness ratios were used to compute the asymmetries at four different intensity regions ([45-60 dB SPL], [55-70 dB SPL], [65-80 dB SPL] and [75-90 dB SPL]) that cover the ranges examined in the present experiments. Asymmetries obtained using these short-term and long-term loudness predictions are reported in Table 1.1. Asymmetries evaluated with DLM and TVL short-term loudness (STL) lie between 0.4 dB and 0.5 dB, which predict very small or almost no difference in loudness between rising and falling tones. Asymmetries predicted with TVL long-term loudness (LTL) are slightly greater (around 1.3 dB) but still fall well below the loudness differences assessed psychophysically. These asymmetries can be seen with TVL time-series on Figure 1.6 where both STL and LTL of a rising-intensity tone reach higher peaks than for a falling-intensity tone.
Note that these predictions are the consequence of the temporal integration stages employed in the models. Indeed, instantaneous loudness patterns are symmetrical for the two profiles. For example, Glasberg and Moore model (2002) uses two successive Automatic Gain Control (AGC) circuits to convert the instantaneous loudness to a short-term loudness pattern first, and then to a long-term loudness pattern. Moreover, while smaller perceptual asymmetries were found for the ramps in the highest regions of our experiments, the models give virtually identical predictions for the different levels (cf. Tab. 1.1).
Preliminary experiment: Individual tone/noise loudness equalization
As stated above, the aim of this second experiment was to evaluate the influence of the intensity region and the spectral content of the sounds on their asymmetries while controlling the potential influence of intrinsic loudness differences between tones and noises. To determine intensity-regions for noises that induced the same loudness as those for tones, preliminary measurements were conducted to evaluate for each participant the exact levels of white noise producing the same loudness levels as 65 and 85-dB SPL 1-kHz tones, that is, only obtained from the maximum intensity of each region. These individualized “equal-loudness” regions (also covering a 15-dB variation) were inferred from a tone/noise loudness-matching task conducted with each participant before the main experiment. The method employed was similar to the main experiment (see section 2.3.1.5 below). An adaptive 2I, 2AFC procedure with four interleaved tracks was used, comprising the two orders of presentation (tone-noise / noise-tone) and two repeated tracks for each order. Each participant matched in loudness 500-ms constant-intensity white noises with 500-ms constant-intensity 1-kHz tones. Tones were presented either at 65 dB SPL or at 85 dB SPL in distinct blocks. The levels of white noises producing the same loudness as 65 and 85-dB SPL 1-kHz tones (averaged across the four tracks) were saved and used individually in the main experiment to produce intensity-regions of equal-loudness.
Table of contents :
Contents
List of Figures
List of Tables
General Introduction
0.1 What is loudness?
0.1.1 Definition
0.1.2 Spectral aspects of loudness
0.1.3 Temporal aspects of loudness
0.1.4 Loudness of natural sounds in natural environments .
0.1.5 The dimension of loudness investigated in this thesis
0.2 Measuring loudness of time-varying sounds
0.2.1 Generalities
0.2.2 Methodological considerations
0.2.3 Issues related to global loudness measurement
0.2.4 Methods considered in this thesis
0.3 Predicting (time-varying) loudness with current models .
0.3.1 The simplest loudness model: Stevens’ power law .
0.3.2 Loudness models for time-varying sounds
0.3.2.1 Early stages taken from stationary loudness models
0.3.2.2 Late temporal integration stages
0.3.2.3 Computing global loudness
0.4 Context of this thesis
0.4.1 Current issues related to the loudness of time-varying sounds
0.4.1.1 Global loudness of long time-varying sounds xliv
0.4.1.2 Sounds with rising vs. falling amplitude envelopes: Perceptual “asymmetries” at different timescales
0.4.1.4 A novel perspective: Temporal weighting of loudness
0.4.1.5 Investigating intensity coding as a complementary approach
0.4.2 Approach of the present thesis
0.4.3 The LoudNat project
0.4.4 Organization of the manuscript
1 A robust asymmetry in loudness between rising and falling intensity tones
1.1 Introduction
1.2 Experiment 1
1.2.1 Method
1.2.1.1 Participants
1.2.1.2 Apparatus
1.2.1.3 Stimuli
1.2.1.4 Procedure
1.2.1.5 Normalization and data analysis
1.2.2 Results
1.2.2.1 Loudness asymmetry
1.2.2.2 Global context effect
1.2.2.3 Local context effect
1.2.3 Discussion
1.3 Experiment 2
1.3.1 Method
1.3.1.1 Participants
1.3.1.2 Apparatus
1.3.1.3 Stimuli
1.3.1.4 Procedure
1.3.2 Results
1.3.2.1 Time order errors
1.3.2.2 Loudness asymmetries
1.3.3 Discussion
1.4 General Discussion
1.4.1 Predictions from current loudness models
1.4.2 Conclusions and perspectives
Partial synthesis
2 Influences of spectral structure and intensity-region on the loudness asymmetry
2.1 Introduction
2.2 Experiment 1
2.2.1 Materials and method
2.2.1.1 Participants
2.2.1.2 Stimuli
2.2.1.3 Apparatus
2.2.1.4 Procedure
2.2.2 Results
2.2.3 Discussion
2.3 Experiment 2
2.3.1 Materials and method
2.3.1.1 Participants
2.3.1.2 Stimuli
2.3.1.3 Apparatus
2.3.1.4 Preliminary experiment: Individual tone/noise loudness equalization
2.3.1.5 Procedure
2.3.2 Results
2.3.3 Discussion
2.4 General discussion and conclusion
3 How temporal profile characteristics of rising and falling tones shape their global loudness
3.1 Introduction
3.2 Experiment 1
3.2.1 Materials and method
3.2.1.1 Participants
3.2.1.2 Stimuli
3.2.1.3 Apparatus
3.2.1.4 Procedure
3.2.2 Results
3.2.3 Discussion
3.3 Experiment 2
3.3.1 Materials and method
3.3.1.1 Participants
3.3.1.2 Stimuli
3.3.1.3 Apparatus
3.3.1.4 Procedure
3.3.2 Results
3.3.2.1 Analysis A – Loudness of constant tones with durations between 2 and 12 s
3.3.2.2 Analysis B – Global loudness of rising and falling ramps varying at 2.5 dB/s
3.3.2.3 Analysis C – Global loudness of rising and falling ramps varying at 5 dB/s
3.3.2.4 Analysis D – Global loudness of rising ramps varying at 2.5 dB/s and 5 dB/s
3.3.2.5 Analysis E – Global loudness of falling ramps varying at 2.5 dB/s and 5 dB/s
3.3.3 Discussion
3.4 General discussion and conclusion
3.4.1 Summary of the present findings
3.4.2 Predicting global loudness from Glasberg and Moore’s model outputs
3.4.3 Conclusion and perspectives
4 Temporal loudness weights of rising and falling tones
4.1 Introduction
4.2 Materials and method
4.2.1 Subjects
4.2.2 Stimuli
4.2.3 Apparatus
4.2.4 Procedure
4.2.5 Data analysis
4.3 Results
4.3.1 Temporal weights for the flat profile
4.3.2 Temporal weights for the increasing and decreasing profiles
4.3.3 Loudness difference between up-ramps and down-ramps
4.4 Discussion and conclusion
Partial synthesis
General Discussion
5.1 Characterization of the global loudness asymmetry between rising and falling sounds
5.1.1 Apparent robustness and constancy of the phenomenon
5.1.2 Influences of the physical attributes of the sounds .
5.1.2.1 Intensity-region
5.1.2.2 Spectral structure
5.1.2.3 Temporal profile characteristics
5.2 Potential mechanisms underlying the asymmetry
5.2.1 Temporal weighting of loudness
5.2.2 Reduction at high intensity-regions and local context emphasis
5.2.3 Decay of the loudness trace of falling stimuli
5.2.4 A specific circuit for rising tones?
5.2.5 Non-sensory factors
5.3 Global loudness processing of time-varying sounds
5.4 Loudness models for time-varying sounds
5.4.1 Predictions of global loudness asymmetries
5.4.2 Challenges for future loudness models
5.5 Main research perspectives derived from this thesis
5.5.1 Identifying the processes highlighted in this thesis
5.5.2 Assessing the role of memory for falling tones
5.5.3 Disentangling decision criterion effects
5.5.4 Achieving a better understanding of temporal integration processes
5.5.5 Investigating global loudness processing with complex natural stimuli
5.5.6 Determining the neural bases of auditory asymmetries
A Interindividual variability in the size of the asymmetry
B Asymmetries with narrow-band noises
C Correlation analyses of the asymmetries measured in Chapter
D Size of the asymmetries in Chapter 3
E Psychometric functions derived from the complementary experiment in Chapter 4
F Modeling global loudness processing of time-varying sounds: A second-order reverse correlation analysis
F.1 Introduction
F.2 Inferring observers’ behavior from both linear and nonlinear kernels
F.2.1 First-order kernels
F.2.2 Second-order kernels
F.2.3 Learning observed in second-order kernels
F.2.4 System identification through second-order kernel properties
F.3 Modeling
F.3.1 Potential modeling structure
F.3.2 Computations
F.3.3 Limitations of the model
F.4 General discussion
G Temporal weighting of loudness in two different loudnessjudgment tasks
G.1 Introduction
G.2 Experiment 1
G.2.1 Materials and method
G.2.1.1 Participants
G.2.1.2 Stimuli
G.2.1.3 Apparatus
G.2.1.4 Procedure
G.2.1.5 Fitting loudness functions
G.2.1.6 Decisions models
G.2.2 Results
G.2.2.1 Loudness functions
G.2.2.2 Temporal weighting patterns
G.2.2.3 Predictive power of the decision models .
G.2.2.4 Increased predictive power by including temporal weights
G.2.2.5 Additional analyses per mean level in the AME task
G.2.3 Discussion
G.3 Experiment 2
G.3.1 Materials and method
G.3.1.1 Participants
G.3.1.2 Stimuli and apparatus
G.3.1.3 Procedure
G.3.1.4 Fitting loudness functions
G.3.1.5 Decisions models
G.3.2 Results
G.3.2.1 Loudness functions
G.3.2.2 Temporal weighting patterns
G.3.2.3 Predictive power of the decision models .
G.3.2.4 Increased predictive power by including temporal weights
G.3.3 Discussion
G.4 General discussion and conclusion
Bibliography
