fix int16_t bytes normalized to float

This commit is contained in:
jiqiang.fu
2023-11-13 17:14:34 +08:00
parent e30a7e32a9
commit 03fb810fab

View File

@@ -46,10 +46,10 @@ public:
// Call it in predict func. if you prefer raw bytes input. // Call it in predict func. if you prefer raw bytes input.
void bytes_to_float_tensor(const char *pcm_bytes) void bytes_to_float_tensor(const char *pcm_bytes)
{ {
std::memcpy(input.data(), pcm_bytes, window_size_samples * sizeof(int16_t)); const int16_t * in_data = reinterpret_cast<const int16_t*>(pcm_bytes);
for (int i = 0; i < window_size_samples; i++) for (int i = 0; i < window_size_samples; i++)
{ {
input[i] = static_cast<float>(input[i]) / 32768; // int16_t normalized to float input[i] = static_cast<float>(in_data[i]) / 32768; // int16_t normalized to float
} }
} }