audio_duration_expand_token_major.c File Reference

#include "ckernel_audio_duration_expand_token_major.h"
#include "ckernel_tts.h"
#include <stdint.h>

Go to the source code of this file.

Functions
int	audio_duration_expand_token_major_f32 (const float *features, size_t input_elements, size_t channels, size_t phoneme_count, size_t input_stride, const int32_t *durations, size_t expanded_frames, float *output, size_t output_elements, size_t output_stride)

Function Documentation

audio_duration_expand_token_major_f32()

int audio_duration_expand_token_major_f32	(	const float *	features,
		size_t	input_elements,
		size_t	channels,
		size_t	phoneme_count,
		size_t	input_stride,
		const int32_t *	durations,
		size_t	expanded_frames,
		float *	output,
		size_t	output_elements,
		size_t	output_stride
	)

Definition at line 6 of file audio_duration_expand_token_major.c.

                          {
    if (!features || !durations || !output || !channels || !phoneme_count ||
        !expanded_frames || input_stride < channels ||
        output_stride < expanded_frames) return CK_AUDIO_EXTENT_INVALID;
    if (phoneme_count - 1 > (SIZE_MAX - channels) / input_stride ||
        channels - 1 > (SIZE_MAX - expanded_frames) / output_stride)
        return CK_AUDIO_EXTENT_OVERFLOW;
    const size_t required_input = (phoneme_count - 1) * input_stride + channels;
    const size_t required_output = (channels - 1) * output_stride + expanded_frames;
    if (required_input > SIZE_MAX / sizeof(float) ||
        required_output > SIZE_MAX / sizeof(float)) return CK_AUDIO_EXTENT_OVERFLOW;
    if (input_elements < required_input || output_elements < required_output)
        return CK_AUDIO_EXTENT_LIMIT;
 
    size_t sum = 0;
    for (size_t token = 0; token < phoneme_count; ++token) {
        if (durations[token] <= 0) return CK_AUDIO_EXTENT_INVALID;
        if ((size_t)durations[token] > expanded_frames - sum)
            return CK_AUDIO_EXTENT_LIMIT;
        sum += (size_t)durations[token];
    }
    if (sum != expanded_frames) return CK_AUDIO_EXTENT_INVALID;
 
    size_t frame = 0;
    for (size_t token = 0; token < phoneme_count; ++token) {
        const float *row = features + token * input_stride;
        for (int32_t repeat = 0; repeat < durations[token]; ++repeat) {
            for (size_t channel = 0; channel < channels; ++channel)
                output[channel * output_stride + frame] = row[channel];
            ++frame;
        }
    }
    return CK_AUDIO_EXTENT_OK;
}

References CK_AUDIO_EXTENT_INVALID, CK_AUDIO_EXTENT_LIMIT, CK_AUDIO_EXTENT_OK, CK_AUDIO_EXTENT_OVERFLOW, and token.