This repository

As part of my Postdoc project, I collected subjective ratings of speaker and voice characteristics. In this repository, I explore data resulting from listening tests with speech stimuli from the Nautilus Speaker Characterization (NSC) Corpus [1].

With respect to speaker characteristics:

• Factor analysis to find traits of social speaker characteristics (SC), with data from Listening Test 1.

  (Folders: \analysis\speaker_characteristics\explorative_analysis_ratings, \analysis\speaker_characteristics\factor_analysis)

  The resulting dimensions (referred to as "traits") were: warmth, attractiveness, confidence, compliance, maturity.

  "WAAT" stands for "warmth-attractiveness", the first two traits resulting from this factor analysis.

  (Publication: [1])

  ​

• Factor analysis to find dimensions of voice descriptions (VD), with data from Listening Test 2.

  (Folders: \analysis\voice_descriptions\explorative_analysis_ratings, \analysis\voice_descriptions\factor_analysis)

  (Publication: [1])

  ​

• Statistical data analysis of effects of telephone degradations (channel bandwidth) on speaker characteristics, with data from Listening Test 3.

  (Folder: \analysis\speaker_characteristics\effects_telephone_degradations)

  (Publication: [2])

  ​

• Statistical data analysis of effects of telephone degradations (channel bandwidth) on voice descriptions, with data from Listening Test 4.

  (Folder: \analysis\voice_descriptions\effects_telephone_degradations)

  ​

• Analyzing the relationship between speech quality and speakers' WAAT, with data from Listening Test 5.

  (Folder: \analysis\relationships_quality_WAAT)

  (Publication: [3])

  ​

• Analyzing the relationships between subjective speaker characteristics and voice descriptions, with data from Listening Test 1 and Listening Test 2.

  (Folder: \analysis\relationships_SC_VD)

  ​

• Computing feature importance measures for the prediction of speaker characteristics and voice descriptions.

  (Folder: \analysis\features_SC_VD)

  (Publication: [8])

  ​

• Predicting subjective ratings of voice descriptions from speech features (regression).

  (Folder: \analysis\predicting_VD)

  ​

  ​

With respect to voice likability:

• Round-robin experiment to analyze interpersonal perceptions of voice likability and personality, with data from Listening Test 6.

  (Folder: \analysis\voice_likability\round-robin)

  (Publication: [4])

  ​

• Paired-comparison experiment (Listening Test 7) to show that this is a valid approach to collect subjective likability ratings.

  (Folder: \analysis\voice_likability\paired_comparison)

  (Publication: [5])

  ​

• Comparison between speaker likability scores obtained by direct scaling or pairwise comparisons in laboratory environments (Listening Tests 6 and 7) or via crowdsourcing (Listening Tests 8 and 9). We used the Crowdee platform to conduct mobile-crowdsourcing experiments based on micro-tasks.

  (Folder: \analysis\voice_likability\crowdsourcing)

  (Publications: [6, 7] )

  ​

ML repository

I also employed these subjective data to perform predictive modeling of speaker characteristics from speech features (see my "ML_Speaker_Characteristics" repository).

Contributing

You are welcome to contribute to this project in any way. Please feel free to fix any errors or send me any suggestion for improvement. If you work at a research institution, you can get the NSC speech files from here.

References

• See my complete list of project publications here.

• See my slides and posters in conference presentations here.

  ​

[1] Fernández Gallardo, L. and Weiss, B., "The Nautilus Speaker Characterization Corpus: Speech Recordings and Labels of Speaker Characteristics and Voice Descriptions," in International Conference on Language Resources and Evaluation (LREC), 2018.

[2] Fernández Gallardo, L., "Effects of Transmitted Speech Bandwidth on Subjective Assessments of Speaker Characteristics," Int. Conf. on Quality of Multimedia Experience (QoMEX), 2018.

[3] Fernández Gallardo, L., Mittag, G., Möller, S. and Beerends, J., "Variable Voice Likability Affecting Subjective Speech Quality Assessments," Int. Conf. on Quality of Multimedia Experience (QoMEX), 2018.

[4] Fernández Gallardo, L. and Weiss, B., "Speech Likability and Personality-based Social Relations: A Round-Robin Analysis over Communication Channels," Interspeech, pp. 903-907, 2016.

[5] Fernández Gallardo, L. "A Paired-Comparison Listening Test for Collecting Voice Likability Scores," Informationstechnische Gesellschaft im VDE (ITG) Conference on Speech Communication, pp. 185-189, 2016.

[6] Fernández Gallardo, L., Zequeira Jiménez, R. and Möller, S., "Perceptual Ratings of Voice Likability Collected through In-Lab Listening Tests vs. Mobile-Based Crowdsourcing," in Interspeech, pp. 2233-2237, 2017.

[7] Zequeira Jiménez, R., Fernández Gallardo, L. and Möller, S., "Scoring Voice Likability using Pair-Comparison: Laboratory vs. Crowdsourcing Approach," Int. Conf. on Quality of Multimedia Experience (QoMEX), 2017.

[8] Fernández Gallardo, L. and Weiss, B., "Perceived Interpersonal Speaker Attributes and their Acoustic Features," in 13. Tagung Phonetik und Phonologie im deutschprachigen Raum, 2017.