Multi-dimensional Speaker Information Recognition with Multi-task Neural Network

Chen, Hạixia; Xu, Longting; Yang, Zhen

doi:10.1109/compcomm.2018.8780705

Cited by 3 publications

(4 citation statements)

References 9 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…However, on the gender tasks, the results are close between the single task learning and multi-task learning. The work in [2] used the KSUEmtotion dataset for emotion and gender recognition; the accuracy recorded was 79.3% for the task of emotion recognition and 98.7% on gender detection. The best experiment of the proposed model outperformed these results as shown in table 4.…”

Section: Data Specific Resultsmentioning

confidence: 99%

“…However, in this paper, the emotions are categorized following [19]'s categorization: anger, surprise, happiness, sadness, questioning, and neutral. Emotion detection from Arabic speech has been explored in previous work as in [13,2], where authors used multiple wellknown classification methods and ensembles on raw features such as MFCC and pitch. Gender detection is an easier task in comparison with emotion and dialect detection in most languages including Arabic.…”

Section: Arabic Speaker Profilingmentioning

confidence: 99%

“…In [3], authors used multi-task learning for estimating age, height, weight, and smoking habits of speakers, using utterance modeling techniques and other classifiers. Researchers in [2] proposed a multi-task learning approach to simultaneously estimate speaker identity and the other two traits of the speaker, specifically, emotion and gender from Arabic speech. This paper proposes the use of a multi-task learning paradigm to simultaneously estimate the gender, dialect, and emotion of the speaker, which slightly outperformed single task classifiers on the three aforementioned tasks.…”

Section: Multi-task Speaker Profilingmentioning

confidence: 99%

“…Although voice signals are mixed signals from which multiple features can be extracted simultaneously, until now, the majority of the voice-related recognition tasks pay more attention to the identification of a single feature from any voice utterance [1]. By contrast, few recognition systems can identify multiple features such as the speaker's gender and emotion simultaneously [2,3]. The availability of speech recognition systems drastically decreases in the case of the Arabic language in line with the deficiency of available datasets.…”

Section: Introductionmentioning

confidence: 99%

See 3 more Smart Citations

SPARTA: Speaker Profiling for ARabic TAlk

Farhan¹,

Za’ter²,

Obaidah³

et al. 2020

Preprint

View full text Add to dashboard Cite

This paper proposes a novel approach to an automatic estimation of three speaker traits from Arabic speech: gender, emotion, and dialect. After showing promising results on different text classification tasks, the multi-task learning (MTL) approach is used in this paper for Arabic speech classification tasks. The dataset was assembled from six publicly available datasets. First, The datasets were edited and thoroughly divided into train, development, and test sets (open to the public), and a benchmark was set for each task and dataset throughout the paper. Then, three different networks were explored: Long Short Term Memory (LSTM), Convolutional Neural Network (CNN), and Fully-Connected Neural Network (FCNN) on five different types of features: two raw features (MFCC and MEL) and three pre-trained vectors (i-vectors, d-vectors, and x-vectors). LSTM and CNN networks were implemented using raw features: MFCC and MEL, where FCNN was explored on the pre-trained vectors while varying the hyper-parameters of these networks to obtain the best results for each dataset and task. MTL was evaluated against the single task learning (STL) approach for the three tasks and six datasets, in which the MTL and pre-trained vectors almost constantly outperformed STL. All the data and pre-trained models used in this paper are available and can be acquired by the public.

show abstract

Section: Data Specific Resultsmentioning

confidence: 99%

Section: Arabic Speaker Profilingmentioning

confidence: 99%

Section: Multi-task Speaker Profilingmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations