Skip to main content

Audio Compression Module (PCA, OPUS)

To transmit audio data over limited BLE bandwidth, we need to compress the audio before transmission. There are two compression methods: PCA and OPUS.

PCA Compression​

The main interfaces for this method are implemented in cpu\br28\fft_and_pca.c and cpu\br28\fft_and_pca.h. The key functions to pay attention to are:

  • hw_fft_init(): This is the initialization function for the FFT calculation module in the PCA compression algorithm. You must execute this function before performing FFT calculations.
  • pca_task(): Starts a task to perform PCA compression. This function calls the following methods:
    • mic_and_spk_data_pop_and_pre_emphasis(s16* emphasized_mic_and_spk_data, int spk_data_count, float sampling_ratio): Preprocesses (e.g., pre-emphasizes) the audio data obtained from the microphone and speaker (only during calls).
    • dct_transform(s16* input, s16* output): Performs DCT transformation on the data, converting it from the time domain to the frequency domain, utilizing the FFT module.
    • split_sign(s16* input, u8* sign): Separates the phase information from the frequency domain data into amplitude without sign and phase information represented as 0/1 binary.
    • compress_and_add_send_block(s16* spectrum, s8* sign): After PCA compression of the amplitude information, combines it with the sign information and inputs it into the BLE send queue.
  • pca_open(): Performs the above two operations and starts the PCA task.
  • transcription_open() and transcription_close(): Enable/disable the transcription state of the earphones. When transcription is enabled, it can prevent repeatedly opening the microphone during calls and closing it after calls.

OPUS Compression​

The main interfaces for this method are implemented in cpu\br28\audio_common\audio_mic_codec.c and cpu\br28\audio_common\audio_dec_codec.c, which are used to compress audio data from the microphone and speaker (only during calls), respectively. The key functions to pay attention to are:

  • audio_mic_enc_open(int (*mic_output)(void *priv, void *buf, int len), u32 code_type, u8 ai_type) and audio_dec_enc_open(int (*dec_output)(void *priv, void *buf, int len), u32 code_type, u8 ai_type): Enable OPUS transcription for microphone/speaker audio. For specific usage, refer to apps\common\third_party_profile\jieli\JL_rcsp\bt_trans_data\le_rcsp_adv_module.c.
  • audio_mic_enc_close() and audio_dec_enc_close(): Disable OPUS transcription for microphone/speaker audio.