[The application of Hilbert-Huang transform in speech coding algorithm for cochlear implants].

Area of Science:

• Signal Processing
• Acoustics
• Speech Technology

Background:

• Realistic listening environments pose challenges for speech recognition.
• Tonal language and music perception require accurate speech signal representation.
• Existing speech coding strategies may not fully capture speech nuances.

Purpose of the Study:

• To introduce a new speech coding strategy based on the Hilbert-Huang transform.
• To improve speech recognition in noisy environments.
• To enhance the perception of tonal languages and music.

Main Methods:

• Empirical Mode Decomposition (EMD) and Hilbert transform to extract instantaneous frequency and amplitude.
• Synthesis of stimulating pulses reflecting speech content, rhythm, and tones.
• Simulation using MATLAB for continuous interleaved sampling and frequency amplitude modulation encoding.

Main Results:

• The proposed Hilbert-Huang transform-based strategy achieved the highest correlation coefficient.
• Synthesized signals closely matched original Mandarin speech signals.
• The new strategy retained more information compared to two other tested strategies.

Conclusions:

• The novel speech coding strategy effectively enhances speech signal representation.
• This approach shows significant potential for improving speech recognition and perception.
• The method is particularly promising for tonal languages and complex auditory signals.