Abstract
The speech signal contains a vast spectrum of information about the speaker such as speakers’ gender, age, accent, or health state. In this paper, we explored different approaches to automatic speaker’s gender classification and age estimation system using speech signals. We applied various Deep Neural Network-based embedder architectures such as x-vector and d-vector to age estimation and gender classification tasks. Furthermore, we have applied a transfer learning-based training scheme with pre-training the embedder network for a speaker recognition task using the Vox-Celeb1 dataset and then fine-tuning it for the joint age estimation and gender classification task. The best performing system achieves new state-of-the-art results on the age estimation task using popular TIMIT dataset with a mean absolute error (MAE) of 5.12 years for male and 5.29 years for female speakers and a root-mean square error (RMSE) of 7.24 and 8.12 years for male and female speakers, respectively, and an overall gender recognition accuracy of 99.60%.
Highlights
Speech is a multidimensional phenomenon, the production of which consists of many anatomical structures movements that influence the overall speech quality and voice characteristics
On the TIMIT test set it achieves root-mean square error (RMSE) of 7.91 and 7.37 years for female/male and mean absolute error (MAE) of 5.2 and 5.37 years female/male. the results in terms of MAE are comparable with the state-of-the-art age estimation results of 4.23 and 5.78 female/male on the NIST SRE08 dataset which were published in [21]
The performance of the systems when trained only on the TIMIT train dataset is already comparable to that of the baseline x-vector system in terms of age estimation and better in terms of gender classification accuracy. This system benefits from pre-training the classifiers and regressor on the Common Voice dataset, as it yields an improvement of 0.56 years RMSE when compared to the system trained without the pre-training
Summary
Speech is a multidimensional phenomenon, the production of which consists of many anatomical structures movements that influence the overall speech quality and voice characteristics. The systems for automatic extraction of these information from speech might be very useful in numerous applications such as in personal identification in banking systems; customer care applications such as call centers; voice bots; and interactive, intelligent voice assistants. Extracting information about age and gender of the speaker may be used by the interactive voice response system (IVR) to redirect the speaker to an appropriate consultant [4] or to play a suitable for a given gender/age group background music [5]. For voice-bots systems, extraction of para-linguistic information may be applied to alter the behaviour of the bot. In the case of voice assistants, such knowledge may be used to target suitable advertisements or select search results that are more fitting for a given age/gender group. All combined, exploiting the para-linguistic content can lead to an improved user experience, and this in turn may generate revenue for the company that decides to use such systems
Sign-in/Register to view highlights & summary Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a callDisclaimer: All third-party content on this website/platform is and will remain the property of their respective owners and is provided on "as is" basis without any warranties, express or implied. Use of third-party content does not indicate any affiliation, sponsorship with or endorsement by them. Any references to third-party content is to identify the corresponding services and shall be considered fair use under The CopyrightLaw.
