Mental health is a growing concern, especially among young adults, but gathering data from this demographic presents distinct challenges. Crowdsensing is a research approach that has become increasingly popular due to its ability to collect data from many individuals continuously and at scale. However, it is equally important to ensure that the data collected is of high quality, as it depends on many factors. In this paper, we discuss the data quality issues encountered during our crowdsensing study conducted from October 2022 to August 2023, which aimed at collecting data to study college students' emotions and mental wellness. We present our findings on data quality issues related to participant recruitment, device usability, data quantity, compliance, consistency, privacy concerns, and incentive mechanisms. We discuss the strategies to address these challenges and plans for future improvements. Our results and discussion highlight the effectiveness of crowdsensing in data collection for this demographic. Additionally, we identified positive and negative emotional drivers and potential stressors affecting this group's mental wellness. The insights from this work can aid the design of future crowdsensing applications and studies.

Research in the field of speech emotion recognition (SER) relies on the availability of comprehensive datasets to make it possible to design accurate emotion detection models.This study introduces the Multimodal Emotion Recognition and Sentiment Analysis (MERSA) dataset, which includes both natural and scripted speech recordings, transcribed text, physiological data, and self-reported emotional surveys from 150 participants collected over a two-week period.This work also presents a novel emotion recognition approach that uses a transformer-based model, integrating pre-trained wav2vec 2.0 and BERT for feature extractions and additional LSTM layers to learn hidden representations from fused representations from speech and text.Our model predicts emotions on dimensions of arousal, valence, and dominance.We trained and evaluated the model on the MSP-PODCAST dataset and achieved competitive results from the best-performing model regarding the concordance correlation coefficient (CCC).Further, this paper demonstrates the effectiveness of this model through crossdomain evaluations on both IEMOCAP and MERSA datasets.