[Low frequency-based non-uniform sampling strategy to improve Chinese recognition in cochlear implant]

Area of Science:

• Audiology and Speech Science
• Biomedical Engineering
• Signal Processing

Background:

• Cochlear implants aim to restore hearing but often struggle with speech and tone recognition, particularly in noisy conditions.
• Existing speech coding strategies like Continuous Interleaved Sampling (CIS) have limitations in capturing crucial tonal information for languages like Mandarin.

Purpose of the Study:

• To introduce and evaluate a novel speech coding strategy, low-frequency fine-structure zero-crossing (LFFS), for cochlear implants.
• To assess the efficacy of LFFS in enhancing Mandarin tone and speech recognition, especially in the presence of background noise.

Main Methods:

• Developed LFFS based on low-frequency non-uniform sampling, utilizing zero-crossing times of the fine structure to generate stimulus pulse sequences.
• Employed a frequency selection rule and an improved index distribution model for stimulus generation.
• Conducted acoustic simulations to compare LFFS performance against CIS under quiet and noisy conditions.

Main Results:

• LFFS demonstrated performance comparable to CIS in quiet environments.
• In noisy backgrounds, LFFS significantly outperformed CIS in recognizing Chinese tones, words, and sentences.
• The improved index distribution model with LFFS resulted in a better distribution of Mandarin recognition factors.

Conclusions:

• LFFS effectively preserves more tonal information, crucial for Mandarin speech comprehension in cochlear implant users.
• The proposed LFFS strategy offers a significant advancement in speech coding for cochlear implants, particularly for tonal languages in challenging acoustic environments.