The focus of this thesis is to investigate the role of voice pitch (the perception of fundamental frequency and/or corresponding harmonics) and apparent vocal tract length (VTL-the perception of the vocal tract resonances i.e. formant frequencies) in perceptions of trustworthiness and other related social attributions. Past research has found that women trust men with relatively higher pitched voices as long-term romantic partners. People with relatively higher pitched voices are also judged as more cooperative than people with relatively lower pitched voices. However, women choose men with relatively lower pitched voices when asked to select which leaders are more trustworthy and make better economic decisions. In study 1, I used “The Trust Game” to determine whether women trust men with higher or lower pitched voices to evenly divide a sum of money. Women trusted men with relatively higher pitched voices more often. Thus, even though men with lower pitched voices are more often elected to office, and are CEOs of larger companies that make more money, women trust men with relatively low pitched voices less than men with relatively high pitched voices to equitably distribute money. Surprisingly, no studies have examined the relationship between VTL and trust, but one recent study examined the relationship between VTL and perceptions of cooperation. In study 2, I was the first to test the role of voice pitch and VTL on perceptions of trust and cooperation. In general, people with higher frequency voices (high pitch and a shorter VTL) were relatively more cooperative and trustworthy than people with lower frequency voices (low pitch and a longer VTL). Despite correlations between the effects of voice frequency manipulations on ratings of trustworthiness and cooperativeness, the amount to which people thought pitch and VTL affected cooperativeness and trustworthiness was different enough to determine that these two constructs overlap, but are not synonymous. Together, these studies show that despite the fact that masculine men tend to win political elections and run large and successful companies, they are viewed as uncooperative and untrustworthy. Future research should investigate if those who win political elections and run successful companies do so because they keep more than their fair share of money. / Thesis / Master of Science (MSc)

One of the main problems faced by automatic speech recognition is the variability of the testing conditions. This is due both to the acoustic conditions (different transmission channels, recording devices, noises etc.) and to the variability of speech across different speakers (i.e. due to different accents, coarticulation of phonemes and different vocal tract characteristics). Vocal tract length normalisation (VTLN) aims at normalising the acoustic signal, making it independent from the vocal tract length. This is done by a speaker specific warping of the frequency axis parameterised through a warping factor. In this thesis the application of VTLN to multiparty conversational speech was investigated focusing on the meeting domain. This is a challenging task showing a great variability of the speech acoustics both across different speakers and across time for a given speaker. VTL, the distance between the lips and the glottis, varies over time. We observed that the warping factors estimated using Maximum Likelihood seem to be context dependent: appearing to be influenced by the current conversational partner and being correlated with the behaviour of formant positions and the pitch. This is because VTL also influences the frequency of vibration of the vocal cords and thus the pitch. In this thesis we also investigated pitch-adaptive acoustic features with the goal of further improving the speaker normalisation provided by VTLN. We explored the use of acoustic features obtained using a pitch-adaptive analysis in combination with conventional features such as Mel frequency cepstral coefficients. These spectral representations were combined both at the acoustic feature level using heteroscedastic linear discriminant analysis (HLDA), and at the system level using ROVER. We evaluated this approach on a challenging large vocabulary speech recognition task: multiparty meeting transcription. We found that VTLN benefits the most from pitch-adaptive features. Our experiments also suggested that combining conventional and pitch-adaptive acoustic features using HLDA results in a consistent, significant decrease in the word error rate across all the tasks. Combining at the system level using ROVER resulted in a further significant improvement. Further experiments compared the use of pitch adaptive spectral representation with the adoption of a smoothed spectrogram for the extraction of cepstral coefficients. It was found that pitch adaptive spectral analysis, providing a representation which is less affected by pitch artefacts (especially for high pitched speakers), delivers features with an improved speaker independence. Furthermore this has also shown to be advantageous when HLDA is applied. The combination of a pitch adaptive spectral representation and VTLN based speaker normalisation in the context of LVCSR for multiparty conversational speech led to more speaker independent acoustic models improving the overall recognition performances.

No description available.

The human voice may convey meaningful information about socially and evolutionarily relevant characteristics of the vocalizer. In turn, listeners may readily evaluate personal characteristics, such as body size, on the basis of nonverbal voice features. Research investigating vocal communication of physical size in humans has focused on two salient and largely independent voice features, fundamental frequency and/or corresponding harmonics (perceived as voice pitch) and formant frequencies (resonance frequencies of the supralaryngeal vocal tract). However, the degree to which fundamental and formant frequencies reliably predict variation in body size controlling for sex and age, and their relative role in the perception or accurate estimation of body size, has to date been unclear. In the current thesis, using meta-analysis, I establish that formants reliably predict variation in men’s and women’s heights and weights. In contrast, fundamental frequency only weakly predicts men’s heights and women’s weights. These findings corroborate work on many other mammals whose vocal production, like humans, follows the source-filter model. Despite the lack of a robust physical relationship between fundamental frequency and size within sexes, I further demonstrate that listeners utilize voice pitch to accurately gauge men’s relative height. My research suggests that voice pitch indirectly facilitates accurate size assessment by providing a carrier signal (i.e., dense harmonics) for formants. This is the first evidence that pitch does not confound accurate size estimation. Finally, I demonstrate that voices with lowered pitch, but not raised pitch, are perceived as larger when projected from a low than high spatial location. These results suggest that strong cross-modal perceptual biases linking low pitch to low elevation and large size may, in some contexts, cause errors in size estimation. Taken together, this thesis provides a detailed account of human vocal communication of body size, which can play a meaningful role in sexual and social contexts. / Dissertation / Doctor of Philosophy (PhD)