Development and validation of the Brazilian voice bank for various emotions (EMOVOX-BR)

Lima, Héryka Maria Oliveira; Almeida, Larissa Nadjara; Aguiar, Alexandra Christine de; Almeida, Anna Alice

doi:10.1590/2317-1782/e20250111pt

Tabla de contenido Anterior Actual Siguiente

Tabla de contenido

Original Article • CoDAS 38 (3) • 2026 • https://doi.org/10.1590/2317-1782/e20250111en linkcopy

Development and validation of the Brazilian voice bank for various emotions (EMOVOX-BR)

Authorship SCIMAGO INSTITUTIONS RANKINGS

Purpose To develop and validate the Voice Bank for Various Emotions for Brazilian Portuguese (EMOVOX-BR).

Methods Observational and cross-sectional study. The corpus of this study consisted of 1,638 sound signals, in different speech tasks, produced by professional actors and actors in training, native speakers of Brazilian Portuguese (PT-BR). From these audios, those containing the phrase in PT-BR “look at the blue plane” in the variation of the six basic emotions plus neutral emission were selected. In the validation stage, the sample was composed of Brazilian speech-language pathologist judges, with experience in the area of voice, to perform the auditory-perceptual judgment of the voices to select the sound signals to compose and validate the EMOVOX-BR. They judged the identification and valence of the emotion, and which vocal parameters were most decisive in the recognition of emotions. Tests were used to verify intra- and inter-judge agreement and reliability.

Results EMOVOX-BR was made up of 39 audios, 24 male and 15 female voices. In the validation phase, all audios obtained a high accuracy rate in recognizing emotions from voice. The emotions anger, disgust and neutral were the most easily identified, with rates above 70%. The pitch and loudness parameters were the most relevant for recognizing emotions.

Conclusion EMOVOX-BR is a pioneering voice bank in PT-BR, made up of 39 audios from native speakers, with variations in the six basic emotions and neutral emission.

Keywords:
Voice; Voice Recognition; Communication; Emotions; Behavior; Database

Variables	Frequency	%
Sample
Sex
Male	15	57.6%
Female	11	42.3%
Education
Incomplete primary school	0	0%
Complete primary school	0	0%
High school	2	7.6%
Incomplete higher education	19	73%
Complete higher education	4	15.3%
Post-Graduation	1	3.8%
Profession
Professional actors	16	61.53%
Performing Arts students	10	38.46%
Participation of theatrical company
Yes	13	50%
No	13	50%
Years of work
0-2 years	8	30.7%
3-8 years	6	23%
9-12 years	9	34.6%
More than 12 years	3	11.5%
Brazilian region
Southeast	11	42.31%
Northeast	8	30.77%
South	3	11.54%
North	2	7.69%
Midwest	2	7.69%
Audios
Sex
Male	24	61.5%
Female	15	38.4%
Emotion
Surprise	8	20.5%
Sadness	7	17.9%
Anger	7	17.9%
Neutral	7	17.9%
Fear	5	12.8%
Happiness	3	7.6%
Disgust	2	5.4%

Variables	Frequency	Percentage
Education
Graduation	5	50%
MSc	0	0%
PhD	0	0%
Postdoctorate	5	50%
Years since graduation
Less than 1 year	0	0%
1 - 5 years	4	40%
6 - 10 years	0	0%
More than 10 years	6	60%
Specialization in the voice area
Yes	6	60%
No	4	40%
Years working in the voice area
Less than 1 year	0	0%
1 - 5 years	4	40%
6 - 10 years	0	0%
More than 10 years	6	60%
Brazilian region
Southeast	5	50%
Northeast	4	40%
South	1	10%
Hearing disorder
Yes	0	0%
No	10	100%

Variables	Sex	Emotion	KAPPA concordance value
Audio 05	Male	Surprise	0.855
Audio 07	Male	Neutral	0.799
Audio11	Male	Anger	0.711
Audio 12	Male	Surprise	0.808
Audio 14	Male	Neutral	0.799
Audio 15	Male	Happiness	1.0
Audio 18	Male	Anger	1.0
Audio 21	Male	Neutral	1.0
Audio 26	Female	Surprise	0.714
Audio 28	Female	Neutral	0.835
Audio 30	Male	Fear	0.855
Audio 33	Male	Surprise	0.753
Audio 38	Female	Sadness	1.0
Audio 45	Male	Sadness	0.879
Audio 52	Male	Sadness	1.0
Audio 54	Male	Surprise	0.902
Audio 56	Male	Neutral	0.822
Audio 80	Male	Sadness	0.855
Audio 81	Male	Anger	0.899
Audio 82	Male	Surprise	0.783
Audio 105	Male	Neutral	0.877
Audio 121	Female	Fear	1.0
Audio 123	Female	Anger	0.799
Audio 127	Male	Happiness	0.835
Audio 128	Male	Fear	0.893
Audio 132	Male	Disgust	0.711
Audio 137	Male	Anger	0.714
Audio 138	Male	Surprise	0.783
Audio 139	Male	Disgust	1.0
Audio 142	Female	Fear	0.713
Audio 149	Female	Fear	0.855
Audio 150	Female	Sadness	1.0
Audio 151	Female	Anger	0.799
Audio 154	Female	Neutral	0.825
Audio 157	Female	Sadness	1.0
Audio 165	Female	Anger	1.0
Audio 178	Female	Sadness	0.714
Audio 180	Female	Surprise	0.808
Audio 141	Female	Happiness	0.771

Emotion	Frequencies	Success Percentage	P-Value
Neutral	7	84.4%	0.44*
Anger	7	82.6%
Disgust	2	75%
Surprise	7	73.8%
Happiness	3	73.3%
Sadness	7	72.5%
Fear	5	71.3%

Variables	Emotion														P-value
	Happiness		Fear		Sadness		Anger		Surprise		Disgust		Neutral
	n	%	n	%	n	%	n	%	n	%	n	%	n	%
Valency
Positive	27	90%	5	8.3%	1	1.3%	2	2.9%	70	85.7%	4	20%	2	2.9%	0.0001*
Negative	3	10%	51	85%	61	76.3%	63	90%	9	11.3%	16	80%	12	17.1%	0.0001*
Neutral	0	0%	4	6.7%	18	22.5%	5	7.1%	1	1.3%	0	0%	56	80%	0.0001*
Voice and Speech parameters
Pitch	21	70%	30	51.7%	49	61.3%	40	57.1%	62	77.5%	13	65%	36	51.4%	0.003*
Loudness	11	36.7%	28	46.7%	28	35%	35	50%	31	38.8%	5	25%	17	24.3%	0.003*
Articulation	4	13.3%	4	6.7%	3	3.8%	17	24.3%	10	12.5%	4	20%	5	7.1%	0.0002*
F Speed	11	36.7%	18	30%	33	41.3%	33	47.1%	28	36.3%	70	35%	24	34.3%	0.004*

location_on

Sociedade Brasileira de Fonoaudiologia Al. Jaú, 684, 7º andar, 01420-002 São Paulo - SP Brasil, Tel./Fax 55 11 - 3873-4211 - São Paulo - SP - Brazil
E-mail: revista@codas.org.br

rss_feed Acompañe los números de esta revista en su lector de RSS

[1] Pearson’s chi-square test

[TFN100] ^*
Significance p<0.05