Signal Processing and Soft Computing Techniques Driven Analysis of Chhattisgarhi Dialects Using MFCCs

Chhattisgarhi, spoken by over 12 million people in Chhattisgarh, is the most prominent, holding official language status and serving as a cultural and linguistic identifier. Baheliya and Chamari, though less documented, are vital to the region's linguistic diversity. This research investigates the acoustic and phonetic characteristics of three central Indian dialects—Chhattisgarhi, Baheliya, and Chamari—using advanced machine learning techniques and robust feature extraction methods. Leveraging Mel-Frequency Cepstral Coefficients (MFCCs) for feature extraction, this study examines statistical parameters such as mean, median, and variance to distinguish the spectral properties of these dialects. Our contribution includes a detailed analysis of dialectal variations, revealing that Chamari exhibits the highest mean MFCC1 value (0.223), indicative of its higher spectral energy, while Chhattisgarhi shows the most negative mean MFCC2 (-0.322), reflecting unique spectral distinctions. Variance analysis highlights Baheliya and Chhattisgarhi as having greater spectral diversity, particularly in MFCC1 and MFCC2, showcasing their phonetic variability. These findings underline the acoustic uniqueness of each dialect, bridging gaps in existing research on underrepresented dialects. This work contributes to the preservation of linguistic diversity and provides a strong foundation for designing ASR systems tailored to regional dialects, addressing critical challenges in multilingual speech recognition.