Update sound classification JAVA reference app DO NOT SUBMIT #296
Conversation
wangtz
requested review from
jdduke,
khanhlvg,
lintian06 and
yyoon
as code owners
wangtz
changed the title
|
|
||
| import java.nio.FloatBuffer; | ||
|
|
||
| public class AudioBuffer { |
| return buffer.length; | ||
| } | ||
|
|
||
| public void feed(float v) { |
| } | ||
| } | ||
|
|
||
| public float[] getArray() { |
| } | ||
|
|
||
| // TODO: what's the correct name? | ||
| public int feed(float[] data, int size) { |
| } | ||
|
|
||
| private float pcm16ToFloat(short v) { | ||
| return (float) v / 32768; |
There was a problem hiding this comment.
This looks like normalization. So the value of float audio buffer will not exceed 1.0f?
| } | ||
|
|
||
| // TODO: ownership | ||
| public FloatBuffer GetAudioBufferInFloat() { |
There was a problem hiding this comment.
getFloatBuffer to be consistent with TensorBuffer
|
|
||
| public void feed(float[] data, int size) { | ||
| for (int i = 0; i < size; i++) { | ||
| feed(data[i]); |
There was a problem hiding this comment.
This will be very slow. Need to optimize it in the real implementation.
| public class AudioBuffer { | ||
|
|
||
| // TODO: What if the ring buffer is not yet fully filled before invocation? | ||
| private class FloatRingBuffer { |
There was a problem hiding this comment.
Want to double confirm with the ring buffer idea. Does all audio classifier expect the ring buffer input? Are there use cases that only need a single audio clip, i.e. not a ring buffer, but a regular buffer like TensorBuffer?
|
|
||
| public int feed(short[] data, int size) { | ||
| for (int i = 0; i < size; i++) { | ||
| floatRingBuffer.feed(pcm16ToFloat(data[i])); |
There was a problem hiding this comment.
Is using floatRingBuffer determined by the input AudioFormat when creating AudioBuffer? i.e. if the AudioFormat's encoding is ENCODING_PCM_16BIT, will we instantiate a ShortRingBuffer?
No description provided.