From ed490b66cb052c1308117e5e9703d94d8e43239a Mon Sep 17 00:00:00 2001
From: huangyuxin <hyxin2014@126.com>
Date: Tue, 29 Mar 2022 03:20:07 +0000
Subject: [PATCH] update spectrogram, test=asr

---
 examples/aishell/asr1/conf/preprocess.yaml |  5 ++--
 paddlespeech/s2t/transform/spectrogram.py  | 34 ++++++++++++++++------
 2 files changed, 27 insertions(+), 12 deletions(-)

diff --git a/examples/aishell/asr1/conf/preprocess.yaml b/examples/aishell/asr1/conf/preprocess.yaml
index a20ff2ab3..d3992cb9f 100644
--- a/examples/aishell/asr1/conf/preprocess.yaml
+++ b/examples/aishell/asr1/conf/preprocess.yaml
@@ -3,9 +3,8 @@ process:
   - type: fbank_kaldi
     fs: 16000
     n_mels: 80
-    n_frame_length: 25
-    n_frame_shift: 10
-    energy_floor: 0.0
+    n_shift: 160
+    win_length: 400
     dither: 0.1
   - type: cmvn_json
     cmvn_path: data/mean_std.json
diff --git a/paddlespeech/s2t/transform/spectrogram.py b/paddlespeech/s2t/transform/spectrogram.py
index f779b07d4..75787d92b 100644
--- a/paddlespeech/s2t/transform/spectrogram.py
+++ b/paddlespeech/s2t/transform/spectrogram.py
@@ -312,17 +312,33 @@ class IStft():
 
 
 class LogMelSpectrogramKaldi():
-    def __init__(self,
-                 fs=16000,
-                 n_mels=80,
-                 n_frame_length=25,
-                 n_frame_shift=10,
-                 energy_floor=0.0,
-                 dither=0.1):
+    def __init__(
+            self,
+            fs=16000,
+            n_mels=80,
+            n_shift=160,  # unit:sample, 10ms
+            win_length=400,  # unit:sample, 25ms
+            energy_floor=0.0,
+            dither=0.1):
+        """
+        The Kaldi implementation of LogMelSpectrogram 
+        Args:
+            fs (int): sample rate of the audio
+            n_mels (int): number of mel filter banks
+            n_shift (int): number of points in a frame shift
+            win_length (int): number of points in a frame windows
+            energy_floor (float): Floor on energy in Spectrogram computation (absolute)
+            dither (float): Dithering constant
+
+        Returns:
+            LogMelSpectrogramKaldi
+        """
+
         self.fs = fs
         self.n_mels = n_mels
-        self.n_frame_length = n_frame_length
-        self.n_frame_shift = n_frame_shift
+        num_point_ms = fs / 1000
+        self.n_frame_length = win_length / num_point_ms
+        self.n_frame_shift = n_shift / num_point_ms
         self.energy_floor = energy_floor
         self.dither = dither